feat: add a configurable category-based logger (#1352)

A self-respecting server needs good observability which starts with
configurable logging. Llama Stack had little until now. This PR adds a
`logcat` facility towards that. Callsites look like:

```python
logcat.debug("inference", f"params to ollama: {params}")
```

- the first parameter is a category. there is a static list of
categories in `llama_stack/logcat.py`
- each category can be associated with a log-level which can be
configured via the `LLAMA_STACK_LOGGING` env var.
- a value `LLAMA_STACK_LOGGING=inference=debug;server=info"` does the
obvious thing. there is a special key called `all` which is an alias for
all categories

## Test Plan

Ran with `LLAMA_STACK_LOGGING="all=debug" llama stack run fireworks` and
saw the following:


![image](https://github.com/user-attachments/assets/d24b95ab-3941-426c-9ea0-a4c62542e6f0)

Hit it with a client-sdk test case and saw this:


![image](https://github.com/user-attachments/assets/3fee8c6c-986e-4125-a09c-f5dc019682e2)

llama_stack/providers/utils/inference/litellm_openai_mixin.py

@@ -8,6 +8,7 @@ from typing import AsyncGenerator, AsyncIterator, List, Optional, Union
 
 import litellm
 
+from llama_stack import logcat
 from llama_stack.apis.common.content_types import (
     InterleavedContent,
     InterleavedContentItem,
@@ -106,6 +107,8 @@ class LiteLLMOpenAIMixin(
         )
 
         params = await self._get_params(request)
+        logcat.debug("inference", f"params to litellm (openai compat): {params}")
+
         # unfortunately, we need to use synchronous litellm.completion here because litellm
         # caches various httpx.client objects in a non-eventloop aware manner
         response = litellm.completion(**params)