🪲 BUG-#68: Request stream_options.include_usage in SSE streaming

[FernandoCelmer]

Description

• pycodeloop/providers/generic.py — Added stream_options: {include_usage: true} to the body built in _stream(), so OpenAI-compatible servers that only emit a usage chunk when explicitly requested (e.g. Ollama) include token counts in streamed responses.
• tests/providers/test_generic.py — Regression test asserting the outgoing request body carries stream_options.include_usage and that the usage values from the final SSE chunk (choices: []) are captured in the result.

Motivation and Context

Streamed responses always showed input_tokens: 0 / output_tokens: 0 against Ollama and other providers that require the client to opt in to usage reporting. The non-streaming path was unaffected.

Closes #68

Types of changes

• Bug fix (change that fixes an issue)
• New feature (change which adds functionality)
• Documentation

Checklist

• I have performed a self-review of my own code
• I have added tests that prove my fix is effective or that my feature works
• I have updated the CHANGELOG
• I have updated the documentation accordingly

[FernandoCelmer]

🔍 Code Review

Code issues found: 2

#	Severity	Comment
1	[Blocking]	stream_options unconditionally overwrites caller-provided value
2	[Blocking]	stream_options.include_usage sent to all SSE providers without compatibility guard