fix(cli): framework build's force option is documented but never read

[blafourcade]

Problem

FrameworkBuildOptions.force is declared and documented as controlling overwrite behaviour:

/** When true, overwrite existing files at canonical flat paths. Only meaningful in flat mode. */
readonly force?: boolean;

Nothing reads it. FrameworkBuildUseCase.execute() destructures only sourceDir and outDir. The field has writers and zero readers.

What actually drives FlatBuildStrategy.checkCollision is a constructor parameter (flat-build-strategy.ts:32, used at lines 273 and 320), threaded through createFrameworkBuildUseCase in deps.ts.

So aidd framework build --flat --force does work, but not through the options object. Anyone debugging a force-related problem would find force being passed to execute() and reasonably conclude it was wired.

Same shape as #464: a documented surface that does not do what it says.

Second problem: the real wiring is untested

Nothing verifies that the user's --force reaches FlatBuildStrategy at all. Making deps.ts ignore the caller's flag breaks no test.

A test that constructs FlatBuildStrategy directly with a literal true does not close this, because it would still pass if deps.ts stopped threading the flag.

Not a data-loss risk

For the record, since it looked like one: runBuild in ensure-built-marketplace-use-case.ts passes force: true unconditionally, which suggested user files could be overwritten. It cannot. That call has exactly two callers, one passing the .aidd/cache/built/<marketplace>/<target> cache dir and one passing a tmpdir() path deleted immediately before the build. Neither is user-authored content, so forcing an overwrite there is cache invalidation — and the flag was inert anyway.

Expected

• the dead option removed, so the type surface stops claiming a capability it does not have;
• a test at the factory level pinning both sides of the real flag: FlatTargetExistsError without --force, overwrite with it;
• a test pinning that runBuild's outDir stays inside the aidd cache or tmp, so pointing it at a user directory fails loudly.

[blafourcade]

Fixed by #557, merged into next at 90b602c.

The dead force option is removed, a factory-level test pins both sides of the real flag (FlatTargetExistsError without --force, overwrite with it), and a second test pins that runBuild's outDir stays inside the aidd cache or tmp.

Closing manually: the PR targeted next rather than the default branch, so the Closes keyword did not fire.