feat: add T-I-F reliability helpers (TifScore, EvaluateTif)

[ferhimedamine]

Summary

Part of T-I-F RFC Phase 3 — adds type-safe T-I-F reliability helpers.

Changes:

• TifScore struct and TifClassification string constants in types.go
• ComputeTifScore(*FeedbackHistoryResponse) TifScore function
• TifScoreFromMetadata(map[string]interface{}) (TifScore, bool) function
• EvaluateTif(ctx, memoryID) method on Client in client.go
• 12 tests in tif_test.go

Go idioms: Uses string-typed TifClassification with named constants (TifConfidentReuse, etc.) following existing FeedbackSignal pattern in the codebase.

🤖 Generated with Claude Code

---

Reviewed-by: Jean-Sébastien Beaulieu (@SeCuReDmE-main-dev) — T-I-F contract parity review

[SeCuReDmE-main-dev]

Phase 3 SDK review note from the RFC side.

The Go API direction is right: TifScore, TifClassification, ComputeTifScore, TifScoreFromMetadata, and EvaluateTif() match the helper pattern from the other SDKs.

I would keep this draft blocked for two reasons:

1. PR state is currently CONFLICTING and there are no checks reported yet.
2. The T-I-F math must be aligned with MCP and the other SDKs before merge/release. This implementation appears to use raw feedback proportions, while dakera-mcp#123 currently adds thin-evidence base indeterminacy for feedback_count < 3. No-feedback classification also needs to be identical everywhere.

Suggested acceptance gate: add shared golden vectors across Python/JS/Rust/Go/MCP and verify the exact same truth, indeterminacy, falsity, classification, and feedback_count for each vector.

Once the conflict is resolved, checks are green, and parity is locked, this should be a good Go-side Phase 3 helper.

[SeCuReDmE-main-dev]

@ferhimedamine final Phase 3 review complete from my side for the Go SDK PR.

I rechecked the current PR state and the DAK-6566 parity fixes. The previous blockers are resolved:

• CI is green.
• PR is mergeable.
• no-feedback now maps to ask_clarification.
• thin-evidence base indeterminacy is present.
• the 8 golden vectors are present and cover the canonical T-I-F v1 contract.
• 3 downvote + 3 flag correctly prioritizes surface_contradiction.
• metadata reliability parsing remains backward compatible with Phase 1 / Phase 2.

This is review-ready from my side. No further requested changes from me.