Model picker filtering and bulk selection

[Imagine-Brian]

Summary

I implemented filtering and bulk-selection controls for the provider model picker.

The implementation is available in my fork:

• Branch: https://github.com/Imagine-Brian/deepseek-harness/tree/feat/model-picker-filter-bulk-selection
• Commit: Imagine-Brian@4cfe0c3

Motivation

An OpenAI-compatible provider may expose dozens of models. The existing picker requires selecting each model individually and does not provide a way to narrow the list to a particular model family.

Behavior

The change adds:

• Case-insensitive model ID filtering
• Trimming of leading and trailing search whitespace
• Select all for currently visible candidates
• Invert selection for currently visible candidates
• Preservation of hidden candidates' selection state
• A no-matches message when the filter returns no candidates
• English and Chinese localization
• Updated English and Chinese documentation

Bulk actions intentionally affect only visible candidates. This allows users to filter for a model family and select exactly those matching models without changing hidden rows.

Validation

• Targeted component test: 85 tests passed
• Staged lint: passed
• Translation pairing checks: passed
• Whitespace checks: passed

The implementation includes component tests for filtering, no-match behavior, selecting all candidates, inverting visible candidates, and preserving hidden selection state.

Request for feedback

The contribution guide indicates that external pull requests are not currently accepted, so I am sharing the implementation here for feedback and possible adoption by the maintainers.