feat(storage): Add MySQL backed target signature store

[manjari25]

Why?

Signatures are expensive to recompute, so the analyzer needs a durable source of truth keyed by batch. This PR implements the TargetSignatureStore contract from the previous PR on MySQL and exposes it through the orchestrator's storage aggregate.

What?

• Adds the target_signature table:
  • Primary key (queue, batch_id), with no secondary index.
  • A row is written once and never updated.
  • targets is a LONGBLOB, because the largest signatures can be large.
• Adds a queue-bound MySQL TargetSignatureStore:
  • A duplicate create maps to ErrAlreadyExists.
  • A signature whose queue differs from the bound queue is rejected.
  • Get reports a missing batch as found=false, not as an error.
• Adds a small binary codec for the targets: an 8-byte big-endian ID followed by a signed varint distance per entry. The codec lives next to the MySQL store because the byte format is a storage detail, so callers only ever see entity.TargetSignature.
• Adds GetTargetSignatureStore() to the orchestrator Storage aggregate only, so the gateway cannot reach it. The mock is regenerated.
• Documents the table in the schema README.
• Updates doc/rfc/service-scoped-extensions.md for the new accessor and table.

Test Plan

./tool/bazel test //submitqueue/orchestrator/extension/storage/mysql:go_default_test
./tool/bazel test //test/integration/submitqueue/extension/storage/mysql/... --test_output=streamed
make check-gazelle
make check-mocks
make lint
CI

Issue