fix(specs): use MAX_SIGNERS in multisig initialize harness

[Stevengre]

Summary

This replaces the shared initialize-multisig harness's hardcoded 11 bounds with spl_token_interface::instruction::MAX_SIGNERS.

Context

Once the runtime-verification path switches to a 3-signer multisig layout, the shared initialize harness must use the same bound as the implementation. Keeping literal 11 in the spec admits out-of-model branches that the proof-only implementation no longer allows.

Red vs Green

Red:

• The runtime-verification implementation constrains multisig inputs to MAX_SIGNERS = 3.
• The shared spec still accepts 1..=11, so impossible branches survive inside test_process_initialize_multisig / test_process_initialize_multisig2.

Green:

• The harness now uses the canonical MAX_SIGNERS constant.
• The spec domain matches the implementation domain on the proof-only path.

References

• Repository reference:
  • spl_token_interface::instruction::MAX_SIGNERS
• Rust standard library: RangeInclusive::contains
  • https://doc.rust-lang.org/std/ops/struct.RangeInclusive.html

[Stevengre]

Folded into #169 so the proof-only multisig implementation/layout change and the shared initialize harness bound stay in one review unit. Closing this PR as redundant.