Allow summarization threshold setting to be a ratio or token count

[bhavyaus]

chat.advanced.summarizeAgentConversationHistoryThreshold previously only accepted an absolute token count. This makes it accept either:

• a ratio of the model's context window — 0 < value <= 1 (e.g. 0.8 to compact at 80%), or
• an absolute token count — value >= 100 (e.g. 60000).

A new resolveSummarizeThresholdTokens() helper resolves the configured value against the effective context window. Ratios automatically track model / context-size changes, so users no longer have to recompute a token number when the window grows.

Values in the ambiguous (1, 100) gap are rejected with a clear error, so a typo like 80 (meant as 80%) fails loudly instead of silently compacting after 80 tokens.

Tests: 4 new specs in summarizeThreshold.spec.ts covering undefined/non-positive → undefined, ratio resolution, absolute passthrough, and rejection of the ambiguous range. All pass.

[Copilot]

Pull request overview

Updates the Copilot extension’s agent history compaction threshold setting to accept either a ratio of the effective model context window or an absolute token count, then resolves it to a concrete token budget at runtime.

Changes:

• Added resolveSummarizeThresholdTokens() to interpret the configured threshold as either (0, 1] ratio or >= 100 tokens, rejecting ambiguous (1, 100) values.
• Updated agent prompt budget calculation to use the resolved token threshold before clamping to the effective context window.
• Added dedicated unit tests and expanded the setting’s UI description to explain the new input formats.

Show a summary per file

File	Description
extensions/copilot/src/extension/intents/node/agentIntent.ts	Adds the threshold-resolution helper and applies it when computing the agent prompt budget.
extensions/copilot/src/extension/intents/node/test/summarizeThreshold.spec.ts	New tests covering unset/non-positive behavior, ratio resolution, absolute passthrough, and ambiguous-range rejection.
extensions/copilot/package.nls.json	Updates the setting description to document ratio vs absolute token usage.

Copilot's findings

• Files reviewed: 3/3 changed files
• Comments generated: 2