doctor must report deploy drift as an issue instead of aborting the health gate

[EtanHey]

Problem

brainlayer doctor calls _check_deploy_drift, which raises BrainLayerAlarm. One drift result aborts the entire doctor run before schema, launchd, writer, socket, and version checks execute.

This made the M1 health gate unusable during an outage; the first check was also a false positive from a packaged Homebrew install.

Reproduction

On BrainLayer 1.5.1:

brainlayer doctor
# deploy_drift alarm
# traceback
# exit 1 before remaining checks

Expected

Doctor should isolate check failures:

• append a fatal/warning DoctorIssue for deploy drift;
• continue running independent checks;
• return a non-zero overall verdict when appropriate without losing the rest of the diagnostic report.

Acceptance

• A deploy-drift finding appears in structured doctor output.
• Later checks still execute and are included.
• Exceptions inside one check are contained and reported with check identity.
• Existing fail-loud exit semantics remain at the aggregate result boundary.