### What happened?

A virtual key or team whose allowed models are restricted to a provider wildcard like `bedrock/*` is incorrectly granted access to models under a **different, unrecognized** provider prefix such as `bedrockz/...`.

`_model_custom_llm_provider_matches_wildcard_pattern` in `litellm/proxy/auth/auth_checks.py` resolves the model's provider via `get_llm_provider` and then re-checks `f"{custom_llm_provider}/{model}"` against the allowed pattern. This is intended for **bare** model names (e.g. `gpt-4o` → `openai/gpt-4o`). But `get_llm_provider("bedrockz/anthropic.claude-3-5-sonnet-20240620")` loosely resolves the provider to `bedrock` **without stripping the `bedrockz/` prefix**, so the re-check runs against `bedrock/bedrockz/anthropic.claude-3-5-sonnet-20240620`, which matches `bedrock/*` — granting access.

### Minimal reproduction (offline, no network/keys)

```python from litellm.proxy.auth.auth_checks import _model_matches_any_wildcard_pattern_in_list

# key/team allowed only "bedrock/*" print(_model_matches_any_wildcard_pattern_in_list( model="bedrockz/anthropic.claude-3-5-sonnet-20240620", allowed_model_list=["bedrock/*"], )) ```

**Actual:** `True` (acc…