* feat(agglayer): process CLAIM notes from a rollup

Support rollup deposits (mainnet_flag=0) in bridge-in verification.
Previously only mainnet deposits were supported and rollup deposits
would panic. Rollup deposits use two-level Merkle proof verification:
first computing the local exit root from the leaf, then verifying it
against the rollup exit root.

Closes #2394

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: extract leading zeros/flag validation from global index helpers

Address review feedback:
- Remove leading zeros assertion and mainnet flag validation from
  process_global_index_mainnet and process_global_index_rollup helpers.
  These are now done once in verify_leaf before branching.
- The helpers now take [rollup_index_le, leaf_index_le] instead of the
  full 8-element global index.
- In process_global_index_rollup, removed unnecessary byte-swap before
  asserting zero (zero is byte-order-independent). This is now moot
  since the mainnet flag is no longer checked in the helper.
- Updated MASM unit tests to match the new helper signatures.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* chore: regenerate agglayer error constants

Include the auto-generated ERR_MAINNET_FLAG_INVALID constant in the
committed errors file to fix the generated files CI check.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: use non-zero indices in rollup test vectors and fix stack bug

Use non-zero leafIndex (2) and indexRollup (5) in the rollup test
vectors to exercise byte-ordering and stack manipulation paths.

This exposed a bug in verify_leaf's rollup branch: the stack
rearrangement after process_global_index_rollup had leaf_index and
rollup_index in the wrong positions, which was masked when both
were zero.

Also restructure the rollup exit tree test helper to use an
SMT-style approach (setLocalExitRootAt) matching the real
PolygonRollupManager.getRollupExitRoot() construction.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: use distinct error messages for mainnet/rollup flag validation

Split ERR_BRIDGE_NOT_MAINNET into two errors:
- ERR_BRIDGE_NOT_MAINNET: "mainnet flag must be 1 for a mainnet deposit"
- ERR_BRIDGE_NOT_ROLLUP: "mainnet flag must be 0 for a rollup deposit"

The previous error text "bridge not mainnet" read backwards when used
in the rollup helper to reject a mainnet flag.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: clarify mainnet flag comment and add ERR_MAINNET_FLAG_INVALID test

- Fix stale comment in verify_leaf that incorrectly referenced
  "stack position 2" for the mainnet flag (it's at position 5)
- Add explicit MASM test for the mainnet flag boolean validation
  (ERR_MAINNET_FLAG_INVALID) which rejects flag values >= 2

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: remove stale constant reference and useless inline test

- Fix comment referencing non-existent MAINNET_FLAG_STACK_POS constant
- Remove test_mainnet_flag_rejects_invalid_value: it inlined the same
  logic as verify_leaf rather than testing verify_leaf itself, so it
  only proved u32lt.2 works on the value 3, not that verify_leaf
  actually performs the check

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: unify GlobalIndex validation and reject invalid flag values

Make validate_mainnet() and validate_rollup() private, exposing only
validate() which checks the mainnet flag is a valid boolean (0 or 1)
before dispatching. This fixes a bug where flag values >= 2 were
incorrectly accepted by validate_rollup() since is_mainnet() returned
false for any non-1 value.

Add mainnet_flag() accessor and a test for invalid flag rejection.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: define memory pointers relative to existing constants

Express SMT_PROOF_LOCAL_EXIT_ROOT_PTR and SMT_PROOF_ROLLUP_EXIT_ROOT_PTR
relative to PROOF_DATA_PTR and each other, rather than hard-coding
absolute values.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: regroup and document memory layout constants

Reorganize constants into clear sections: proof data memory layout
(with address map comment), leaf data, calculate_root locals, and
data sizes. Define GLOBAL_INDEX_PTR relative to
SMT_PROOF_ROLLUP_EXIT_ROOT_PTR for a fully chained layout.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Apply suggestions from code review

* Fix rollup deposit merkle proof stack setup (#2576)

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

* Apply suggestions from code review

Co-authored-by: Alexander John Lee <77119221+partylikeits1983@users.noreply.github.com>

---------

Co-authored-by: Claude (Opus) <noreply@anthropic.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Alexander John Lee <77119221+partylikeits1983@users.noreply.github.com>