Add tj-test-case skill — help create test case / measurements from a diff

Build the second of the two mechanisms flagged as missing in [docs/regular-usage-workflow.md](docs/regular-usage-workflow.md#two-mechanisms-still-needed):

> Something needs to help turn a known-good `A..D` diff into the actual measurement set — proposing `fileCreated`/`grep`/`command` assertions and, most valuably, drafting the `correctness-test` template — for the human to review before it's committed as `E`.

## Scope to nail down
- Input: presumably a commit range (`A..D`) the user points the skill at.
- Output: proposed measurements (fileCreated/grep/command) plus a drafted correctness-test template, per the mechanics in the workflow doc — always for human review before commit, never auto-committed.
- Needs to also handle the Scenario 2 "specification gap" case — rewriting the task instruction itself, not just mining measurements — per the workflow doc's distinction between convention gaps and specification gaps.

Start with docs/regular-usage-workflow.md, especially the “two mechanisms still needed” section and its Scenario 2 distinction. Define how the skill accepts an A..D commit range, proposes fileCreated/grep/command measurements, drafts a correctness-test template, and rewrites the task instruction for specification gaps. Done means all output is presented for human review and nothing is auto-committed.

Written by the indexing model from the issue text.