chore: regenerate pallet weights

[n13]

Summary

Regenerates benchmark weights for every pallet registered in define_benchmarks whose weights the runtime actually consumes, measured on current main (includes #646 vesting).

What changed

Regenerated weights (50 steps / 20 repeats, compiled WASM): wormhole, multisig, reversible-transfers, scheduler, mining-rewards, treasury, vesting — plus balances and frame-system, which were registered and wired into the runtime but missing from the script. regenerate_weights.sh now covers them.

Fixed broken wormhole benchmarks — pre_validate_proof / pre_validate_public_batch_proof panicked against the live runtime: the committed proof fixtures (test-data/*.hex) still baked in the old 10 bps volume fee, while the runtime has charged 4 bps since #632, so production pre-validation rejected the fixture with InvalidVolumeFeeRate. Mock-based tests never caught it because the mock was pinned to 10 bps to match the fixtures. Fixtures regenerated at 4 bps; mock and affected tests aligned.

Hand-augmented weight files preserved — a raw template run wipes these, so the measured bases were merged into the existing hand layers instead:

• wormhole: refreshed the four timing constants. Notable moves: public-batch pre-validation 4.5B → 1.9B ps (cheaper production path after turnstile removal in Remove wormhole turnstile (soundness counters and address reveal) #645), private pre-validation 700M → 900M ps, ZK verify 14.97B → 15.16B / 20.69B → 21.08B ps. Depth-scaling storage tails and invariant tests kept; the public pre-validation floor test now tracks the post-Remove wormhole turnstile (soundness counters and address reveal) #645 cost.
• mining-rewards: measured base 163M → 162M ps; depth pricing and tests kept.
• reversible-transfers: execute_transfer base 110M → 98M ps; live-depth pricing helper and tests kept, other extrinsics take fresh measurements.
• scheduler: kept the V12 audit #162534 explicit Retries write charge in service_task_base (benchmark still doesn't measure it).

regenerate_weights.sh documents these hand layers so future runs don't silently drop them.

Verification

• cargo check --workspace --release clean
• All unit tests pass for every touched pallet (wormhole 68, mining-rewards 29, reversible-transfers 76, scheduler 274, multisig 56, treasury 15, vesting 88, balances, frame-system), including the weight invariant tests
• Benchmarks re-run end-to-end with no failures