fix(pbs): launch unified commit-boost binary on v0.10+

[crypDuck]

Summary

• start-pbs.sh now runs /usr/local/bin/commit-boost pbs when that binary exists, and falls back to /usr/local/bin/commit-boost-pbs for the v0.9.x image that Hyperdrive still defaults to.
• Neither binary present exits 127 with an explicit message instead of a bare exec: not found.
• Config path is unchanged (CB_CONFIG=/cb_config.toml).

Fixes the crash-loop when a user sets the Commit-Boost container tag to v0.10.0 or newer:
/usr/share/hyperdrive/scripts/start-pbs.sh: 15: exec: /usr/local/bin/commit-boost-pbs: not found

Commit-Boost v0.10.0 unified the CLI, PBS, and signer into one binary (commit-boost-client#425). The v0.11.0 PBS image entrypoint is commit-boost pbs (instead of commit-boost-pbs).

This does not bump the default container tag. v0.11.0 also drops the Builder API v2-to-v1 fallback and requests SSZ from relays first, so that should be a separate change.

Test plan

• On a v1.3.0 node with the tag overridden to v0.11.0, replacing the exec with commit-boost pbs stops the exit-127 loop.
• Fresh hyperdrive service start against the default v0.9.2 image still launches commit-boost-pbs.
• Tag override to v0.11.0 launches commit-boost pbs, registers with the configured relays, and stays up.
• MEV-Boost path in the same script is unchanged.