amazon-cloudwatch-agent doesnt respect SIGTERM to flush metrics

**Describe the bug**

I am trying to run the aws-cloudwatch-agent as part of a lambda extension, so that metrics sent by lambda executions are batched.
Basically, I have a go program that runs this command under the hood with some static configuration files that are pasted below
`amazon-cloudwatch-agent -config config.toml -otelconfig config.yaml`
So far, the bootup works. However, the problem is that the agent doesnt seem to flush the metrics when it receives a SIGTERM signal. It just seems to close all extensions and shutdown without flushing any metrics to cloudwatch. This is really problematic for using inside a lambda extension because until a lambda environment shutdown if the flush interval is not reached, metrics are lost completely

Apparently, the telegraf agent understands a SIGTERM or SIGUSR1 and flushes its cache before shutting down.

[This issue](https://github.com/aws/amazon-cloudwatch-agent/issues/170) already goes in this direction.

**What did you expect to see?**
The metrics should be flushed during shutdown

**What did you see instead?**
The metrics are lost

**What config did you use?**
```
[agent]
collection_jitter = "0s"
debug = false
flush_interval = "1s"
flush_jitter = "0s"
hostname = ""
interval = "60s"
metric_batch_size = 1000
metric_buffer_limit = 10000
omit_hostname = true
precision = ""
quiet = false
round_interval = false

[inputs]

[[inputs.statsd]]
interval = "10s"
parse_data_dog_tags = true
service_address = ":8125"
[inputs.statsd.tags]
"aws:AggregationInterval" = "10s"

[outputs]

[[outputs.cloudwatch]]
```

config.yaml
```
connectors: {}
exporters:
awscloudwatch:
force_flush_interval: 10s
max_datums_per_call: 1000
max_values_per_datum: 150
namespace: somewhere
region: eu-central-1
resource_to_telemetry_conversion:
enabled: true
extensions: {}
processors: {}
receivers:
telegraf_statsd:
collection_interval: 10s
initial_delay: 1s
timeout: 0s
service:
extensions: []
pipelines:
metrics/host:
exporters:
- awscloudwatch
processors: []
receivers:
- telegraf_statsd
```

Start by reproducing the shutdown behavior with the supplied config.toml and config.yaml, then trace the agent's SIGTERM handling and shutdown sequence. Confirm whether buffered metrics are flushed before termination; done means metrics emitted before SIGTERM reach CloudWatch instead of being lost.

Written by the indexing model from the issue text.