Validate explicit targets in layer row commands (part of #785).

[FredApps]

What and why

Malformed or unknown explicit targets now fail before mutation for layer.delete, layer.duplicate, layer.setProps, layer.template, layer.merge, layer.reverse, layer.hideOthers, layer.outlineOthers, and layer.lockOthers, plus layer.collectInNew.ids and layer.flatten.id. Complete target arrays are validated before the existing work cap is applied.

Omitted defaults, ids precedence, and explicit empty-list behavior are preserved. An empty keep list retains the all-layer effects of the Others commands. This addresses part of #785; it does not cover every selector or close the whole issue.

Implementation and testing were AI-assisted under FredApps.

How it was tested

• All three reported failures were reproduced on synthetic baseline documents.
• 29 focused layer tests passed, including six new regressions.
• All seven local cargo xtask ci gates passed: 5,364 tests passed, zero failed, and four ignored.
• Independent source review cleared the bounded scope.

Tested head: 8d60ff69055d421f3c624543ddcf8ee0d83262b5; tree: fa22be9876b984f8dbd1383141db0ddf6fdf2693; base: 53bc86fdb454ed33cd26cba5e29695ebe0befeff.

Main advanced to 2ac3ba2b866ad9d34b2252dd8024f73e34f4b2b8 without changes in the affected files. Integration with that newer main was not tested. No hosted CI or runtime MCP transport pass is claimed.

Checklist

• cargo xtask ci passes (fmt, clippy, tests, assets, brands, layers, wasm), locally on the tested head above.
• No new panics in shipped code: no new unwrap(), expect(), panic!, unreachable!, todo! or
  unimplemented! outside tests, and no new unchecked indexing on data from files, parameters or
  messages. Errors are returned as Result and shown to the user
  (why and how).
• Clean-room: nothing copied from Adobe products or GPL/AGPL code (AGENTS.md).
• Every new asset has a row in ASSETS.md (no new assets).