dlt tries OAuth in GCP Instead of Service Account Credentials when given Service Account Credentials via credential file

### dlt version

dlt 1.17.1

### Describe the problem

When using GCP service account credentials, referenced either in environment variables or in `secrets.toml` I get this following error.

```
`GcpOAuthCredentials` cannot parse the configuration value provided. The value is of type `str` and comes from the sections `('credentials',)` Value may be a secret and is not shown. Details: The expected representation for `GcpOAuthCredentials` is a string with serialized oauth2 user info and may be wrapped in 'install'/'web' node - depending of oauth2 app type.
```

### Expected behavior

dlt should be using `GcpServiceAccountCredentials`, which currently can be used when explicitly provided within the pipeline as in

```
with open("creds.json", "r") as f:
creds = json.load(f)

service_dict = json.loads(json.dumps(creds))

gcp_credentials = GcpServiceAccountCredentials()
gcp_credentials.parse_native_representation(json.dumps(creds))
```

### Steps to reproduce

set up GCP service account credentials and attempt to use them either in secrets.toml as in

```
[destination.filesystem]
bucket_url = "gs://yourbucket...

[destination.filesystem]
credentials="path/to/your/json"
```

when running a dlt pipeline for filesystem destination

### Operating system

macOS

### Runtime environment

Local

### Python version

3.10

### dlt data source

API

### dlt destination

Filesystem & buckets

### Other deployment details

_No response_

### Additional information

_No response_

Start by tracing how the filesystem destination resolves credentials from environment variables and secrets.toml, comparing GcpOAuthCredentials with GcpServiceAccountCredentials. Reproduce the reported pipeline with a service-account JSON credential file and confirm that the file is accepted without explicit credential parsing in the pipeline.

Written by the indexing model from the issue text.