Add a review rule for specs that cannot fail usefully

[bockets]

Adds a shared rule block, prompts/_shared/spec-value-rules.md, asking the review to judge the specs a diff adds or changes.

The definition: a spec is useless if its failure would neither reveal a regression worth fixing nor point at what to fix. The first half rules out the spec that cannot go red. The second rules out the spec that goes red on everything and explains nothing.

Two deliberate limits:

• It never blocks. A useless spec is reported and the PR is still approved unless another finding blocks. Test value is a judgment call.
• It carries a do-not-flag list. Regression specs for fixed bugs, edge and nil-path examples, the required flag-on/flag-off pair, and missing coverage are all named as not this finding. Without that, the likely failure is the review asking to delete specs that look redundant and are not.

Wired into all three prompts so the stages cannot drift, which bumps each prompt's content-derived version.

Same definition proposed for the app repo's CLAUDE.md: BiggerPockets/biggerpockets#30942