inputs.cisco_telemetry_mdt: 'embedded_tags' generating spurious messages

### Relevant telegraf.conf

```toml
[[inputs.cisco_telemetry_mdt]]
transport = 'grpc'
service_address = ':57000'
max_msg_size = 100000000
embedded_tags = [
'ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics/meter-statistics/meter-id'
]

[[outputs.file]]
files = ['/tmp/metrics.log']
data_format = 'json'
json_timestamp_units = '1s'
```

### Logs from Telegraf

```text
-
```

### System info

Telegraf 1.34.3, RockyLinux 9

### Docker

_No response_

### Steps to reproduce

1. Set up a MDT subscription in a Cisco Catalyst device:
```
telemetry ietf subscription 100
encoding encode-kvgpb
filter xpath /if:interfaces-state/interface/target:diffserv-target-entry/diffserv-target-classifier-statistics
source-address 10.0.0.15
source-vrf clab-mgmt
stream yang-push
update-policy periodic 500
no receiver ip address 192.168.100.11 57000 protocol grpc-tcp
receiver ip address 192.168.100.20 57000 protocol grpc-tcp
```

2. Configure some example policies:
```
policy-map INPUT_3Mb
class class-default
police cir 3000000 bc 62500
conform-action transmit
exceed-action drop
policy-map OUTPUT_1Mb
class class-default
shape average 1000000 8000 8000
interface GigabitEthernet2
service-policy output OUTPUT_1Mb
interface GigabitEthernet3
service-policy input INPUT_3Mb
```

### Expected behavior

Receive a couple of MDT metrics like these. Please note `meter_statistics/meter_id` tag in the second metric.

```
{
"fields": {
"classifier_entry_statistics/classified_bytes": 21649700,
"classifier_entry_statistics/classified_pkts": 14530,
"classifier_entry_statistics/classified_rate": 509,
"queuing_statistics/drop_bytes": 10760780,
"queuing_statistics/drop_pkts": 7222,
"queuing_statistics/output_bytes": 10888920,
"queuing_statistics/output_pkts": 7308,
"queuing_statistics/queue_size_bytes": 0,
"queuing_statistics/queue_size_pkts": 64,
"queuing_statistics/wred_stats/early_drop_bytes": 0,
"queuing_statistics/wred_stats/early_drop_pkts": 0
},
"name": "ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics",
"tags": {
"classifier_entry_name": "class-default",
"direction": "outbound",
"name": "GigabitEthernet2",
"parent_path": "OUTPUT_1Mb class-default",
"path": "ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics",
"policy_name": "OUTPUT_1Mb",
"source": "switch",
"subscription": "100"
},
"timestamp": 1727266737
}

{
"fields": {
"classifier_entry_statistics/classified_bytes": 111006574,
"classifier_entry_statistics/classified_pkts": 74503,
"classifier_entry_statistics/classified_rate": 2603,
"meter_statistics/meter_failed_bytes": 0,
"meter_statistics/meter_failed_pkts": 0,
"meter_statistics/meter_succeed_bytes": 21622964,
"meter_statistics/meter_succeed_pkts": 14514,
"queuing_statistics/drop_bytes": 0,
"queuing_statistics/drop_pkts": 0,
"queuing_statistics/output_bytes": 0,
"queuing_statistics/output_pkts": 0,
"queuing_statistics/queue_size_bytes": 0,
"queuing_statistics/queue_size_pkts": 0,
"queuing_statistics/wred_stats/early_drop_bytes": 0,
"queuing_statistics/wred_stats/early_drop_pkts": 0
},
"name": "ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics",
"tags": {
"classifier_entry_name": "class-default",
"direction": "inbound",
"meter_statistics/meter_id": "0",
"name": "GigabitEthernet3",
"parent_path": "INPUT_3Mb class-default",
"path": "ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics",
"policy_name": "INPUT_3Mb",
"source": "switch",
"subscription": "100"
},
"timestamp": 1727266737
}
```

### Actual behavior

The expected metrics are received, but also an additional dummy one for each policy containing a meter id (i.e, the field used for `embedded_tags`). In this example, just one additional dummy metric:

```
{
"fields": {
"classifier_entry_statistics/classified_bytes": 111006574,
"classifier_entry_statistics/classified_pkts": 74503,
"classifier_entry_statistics/classified_rate": 2603,
},
"name": "ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics",
"tags": {
"classifier_entry_name": "class-default",
"direction": "inbound",
"name": "GigabitEthernet3",
"parent_path": "INPUT_3Mb class-default",
"path": "ietf-interfaces:interfaces-state/interface/ietf-diffserv-target:diffserv-target-entry/diffserv-target-classifier-statistics",
"policy_name": "INPUT_3Mb",
"source": "switch",
"subscription": "100"
},
"timestamp": 1727266737
}
```

### Additional info

If `embedded_tags` is not used, the expected number of messages is received. It looks like some kind of bug in the Telegraf MDT plugin when handling the `embedded_tags` option.

This is the Yang model used:

```
$ pyang -f tree ietf-diffserv-target.yang
module: ietf-diffserv-target

augment /if:interfaces/if:interface:
+--rw diffserv-target-entry* [direction policy-name]
+--rw direction identityref
+--rw policy-name string
augment /if:interfaces-state/if:interface:
+--ro diffserv-target-entry* [direction policy-name]
+--ro direction identityref
+--ro policy-name string
+--ro diffserv-target-classifier-statistics* [classifier-entry-name parent-path]
+--ro classifier-entry-name string
+--ro parent-path string
+--ro classifier-entry-statistics
| +--ro classified-pkts? uint64
| +--ro classified-bytes? uint64
| +--ro classified-rate? uint64
+--ro meter-statistics* [meter-id]
| +--ro meter-id uint16
| +--ro meter-succeed-pkts? uint64
| +--ro meter-succeed-bytes? uint64
| +--ro meter-failed-pkts? uint64
| +--ro meter-failed-bytes? uint64
+--ro queuing-statistics
+--ro output-pkts? uint64
+--ro output-bytes? uint64
+--ro queue-size-pkts? uint64
+--ro queue-size-bytes? uint64
+--ro drop-pkts? uint64
+--ro drop-bytes? uint64
+--ro wred-stats
+--ro early-drop-pkts? uint64
+--ro early-drop-bytes? uint64
```

Start by locating the inputs.cisco_telemetry_mdt plugin and its embedded_tags handling. Reproduce the issue with the supplied telegraf.conf and Cisco MDT subscription, then compare emitted metrics with and without embedded_tags. Done means the expected metrics remain present without an additional dummy metric for policies containing the embedded tag.

Written by the indexing model from the issue text.