Data Connector metadata import example fails

### Version Information

Server Version:
CLI Version (for CLI related issue):

### Environment

Hasura/graphql-engine:v2.15.1 on a local K8s cluster.

### What is the current behaviour?

The `POST /v1/metada` request succeeding (following https://github.com/hasura/graphql-engine/tree/master/dc-agents#setting-up-data-connector-agents-with-graphql-engine).

### What is the expected behaviour?

400 Bad Request error returned:
```json
{
"code": "unexpected",
"error": "cannot continue due to inconsistent metadata",
"internal": [
{
"definition": "reference",
"name": "data_connector_agent reference",
"reason": "Inconsistent object: Error in Data Connector backend: Connection failure",
"type": "data_connector_agent"
},
{
"definition": "chinook",
"name": "source chinook",
"reason": "Inconsistent object: Data connector named \"reference\" was not found in the data connector backend info",
"type": "source"
}
],
"path": "$.args"
}
```

### How to reproduce the issue?

1. Have `hasura/graphql-engine` running on a K8s pod (which makes Hasura available on http://localhost:8550).
2. `cd graphql-engine\dc-agents` then run `docker-compose up -d`.
3. Send a POST request to http://localhost:8550/v1/metadata with the body from https://github.com/hasura/graphql-engine/tree/master/dc-agents#setting-up-data-connector-agents-with-graphql-engine.
4. See the output.

### Screenshots or Screencast


### Please provide any traces or logs that could help here.

Hasura pod logs:
```
{"detail":{"http_info":{"content_encoding":null,"http_version":"HTTP/1.1","ip":"192.168.65.3","method":"POST","status":400,"url":"/v1/metadata"},"operation":{"error":{"code":"unexpected","error":"cannot continue due to inconsistent metadata","internal":[{"definition":"reference","name":"data_connector_agent reference","reason":"Inconsistent object: Error in Data Connector backend: Connection failure","type":"data_connector_agent"},{"definition":"chinook","name":"source chinook","reason":"Inconsistent object: Data connector named \"reference\" was not found in the data connector backend info","type":"source"}],"path":"$.args"},"request_id":"a96b352d-5a3d-431a-a6da-d93755dce097","request_mode":"error","response_size":467,"uncompressed_response_size":467,"user_vars":{"x-hasura-role":"admin"}},"request_id":"a96b352d-5a3d-431a-a6da-d93755dce097"},"level":"error","timestamp":"2022-12-15T10:35:20.950+0000","type":"http-log"}
```

### Any possible solutions/workarounds you're aware of?

No, I tried replacing `reference` with `experimental` which was suggested when I tried adding the metadata for the DC alongside an existing data source and actions (I also tried with `sqlite` which I saw being used a lot in the codebase but got the same error).
I also tried changing the data sent to match the expected format based on the errors I was getting on Postman, so `table` and `remote_table` are strings rather than `string[]`.
And using `postgres` as the kind and adding to `configuration`:
```json
"connection_info": {
"use_prepared_statements": false,
"database_url": {
"from_env": "SUBSCRIPTIONS_DB_URL"
},
"isolation_level": "read-committed"
}
```

### Keywords

Data connected name, was not found in the data connector backend, v1/metadata, chinook,

Start with the dc-agents setup instructions and the docker-compose configuration, then reproduce the POST /v1/metadata request against the Kubernetes-hosted Hasura instance. Trace the data connector agent connection failure and verify the documented example imports successfully without inconsistent metadata errors.

Written by the indexing model from the issue text.