fix: improve OpenAI-compatible error handling and add test coverage

[codefromthecrypt]

Improves error handling reliability for OpenAI-compatible providers and adds comprehensive test coverage.

I tried using Goose through a gateway which had an incorrect max payload size. The gateway returned a plain text error.

Before, the underlying error was obfuscated by a JSON failure: Error: Request failed: error decoding response body

$ goose run --debug -t 'should I install goose?'
starting session | provider: openai model: qwen3
    logging to /Users/codefromthecrypt/.local/share/goose/sessions/20250819_090611.jsonl
    working directory: /Users/codefromthecrypt/oss/goose
Error: Request failed: error decoding response body
Interrupted before the model replied and removed the last message.
  error: The error above was an exception we were not able to handle.
These errors are often related to connection or authentication
We've removed the conversation up to the most recent user message
- depending on the error you may be able to continue

Now, the error is easier to see:
Error: Request failed: Request failed with status: 413 Payload Too Large

$  ./target/debug/goose run --debug -t 'should I install goose?'
starting session | provider: openai model: qwen3
    logging to /Users/codefromthecrypt/.local/share/goose/sessions/20250819_090540.jsonl
    working directory: /Users/codefromthecrypt/oss/goose
  2025-08-19T02:05:53.160383Z  WARN goose::providers::utils: Provider request failed with status: 413 Payload Too Large. Payload: None. Returning error: RequestFailed("Request failed with status: 413 Payload Too Large")
    at crates/goose/src/providers/utils.rs:113

Error: Request failed: Request failed with status: 413 Payload Too Large
Interrupted before the model replied and removed the last message.
  error: The error above was an exception we were not able to handle.
These errors are often related to connection or authentication
We've removed the conversation up to the most recent user message
- depending on the error you may be able to continue

Changes:

• Add PartialEq derive to ProviderError enum to enable testing
• Refactor handle_status_openai_compat to improve error parsing and reduce complexity
• Enhance authentication error message formatting with cleaner conditional logic
• Add comprehensive test suite covering various error scenarios and edge cases
• Improve function documentation with clear references to OpenAI error codes

Impact:

• More reliable error handling when provider APIs return malformed responses
• Better error messages for authentication failures
• Increased test coverage prevents regressions in error handling logic
• Cleaner code structure improves maintainability

[michaelneale]

thanks Adrian tagging @DOsinga as I know he had designs on making openai and providers more declarative and configurable

will need a good old cargo fmt run of course

[codefromthecrypt]

rebased!

[michaelneale]

running now...

[michaelneale]

get it in when the lights are green!

[codefromthecrypt]

I see 💚