feat(ollama): add generateWithUsage() and export Usage/GenerateResult types

[pyramation]

Summary

Adds generateWithUsage() to OllamaClient that returns token usage metadata alongside content. The Ollama API returns prompt_eval_count and eval_count in responses, which the OllamaAdapter already captures internally — but the legacy generate() API discarded them and returned only the text string.

Changes:

• New generateWithUsage(input, onChunk?) method returns GenerateResult { content, usage, model, stopReason }
• generate() now delegates to generateWithUsage() internally — fully backward compatible
• Exports Usage and GenerateResult interfaces for consumer type safety
• Works for both streaming (usage from final done: true chunk) and non-streaming modes

Motivation: graphile-llm needs actual provider token counts for billing metering (check_billing_quota → LLM call → record_usage with real token counts). Previously had to estimate from text length (~4 chars/token).

Review & Testing Checklist for Human

• Verify generate() still returns string (non-streaming) and void (streaming) — backward compat
• Test generateWithUsage() with a local Ollama instance: confirm usage.input and usage.output are populated from the provider response
• Check that streaming mode with onChunk callback still delivers text deltas AND returns usage in the GenerateResult

Notes

• The OllamaAdapter.stream() was already capturing prompt_eval_count/eval_count from the done: true chunk (lines 464-466). This change just surfaces that data through the public API.
• Usage was changed from interface to export interface.
• GenerateResult.stopReason includes 'toolUse' to match the full AssistantMessage stopReason union.

Link to Devin session: https://app.devin.ai/sessions/2b5a29d83d3f478e8d3d972653b4879c
Requested by: @pyramation

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring