Problem 1: possibly dead branch in __clayered_modify_try_ingest (src/cursor/cur_layered.c)

• When the ingest value for the key is a tombstone or delete-encoded, the branch forces a full update because such a value cannot serve as a modify base.
• A lookup returns WT_NOTFOUND for a deleted key, so the branch is only reachable if the modify skips the lookup on an already-positioned cursor.
• __clayered_modify_ingest skips the lookup only when both WT_CURSTD_KEY_INT and WT_CURSTD_VALUE_INT are set and a positioned remove clears VALUE_INT. No known path positions a cursor on a tombstone with VALUE_INT set.
• Outcome: either prove the branch unreachable and replace it with an assertion or identify the reaching path and cover it with a Python test.

Problem 2: no modify operation in test/suite/test_layered_cursor_stress.py

• The stress workload exercises insert, update, remove, search and iteration on plain, leader and follower layered cursors but never modify.
• Add a modify op, including modify on a just-removed slot and give it non-zero weight in the workload profiles.

Goal:

• The branch in __clayered_modify_try_ingest is either asserted unreachable or tested.