Remove redundant config re-resolution in doctor diagnostics

[nicoloesch]

Summary

collect_doctor_report() was calling get_cdm_context() a second time even though collect_maintenance_info() (called immediately before it) already resolves the same config. This results in three independent config reads for one doctor CLI invocation once the @omop_command decorator's own resolution is counted.

Now threads the already-resolved result through instead of re-reading. collect_maintenance_info()'s own independent resolution is intentionally left as-is (it's a diagnostic health-check whose job is to test config resolution fresh, not a case of "caller already has it") . Its broad except Exception around that resolution is narrowed to avoid mislabeling unrelated bugs as a generic engine-resolution error.

[nicoloesch]

Removing the Liskov Substitution Principle fix that got picked up by ty>=0.0.62 as privately discussed. The respective bug is now reported here:
#41

Pinning ty to 0.0.61 to pass the static typing check.

[gkennos]

the call here still has unbounded except within collect_maintenance_info - noted since the PR is aiming to clean up the broad exceptions here

maybe (in cli_schema_info.py)

except (FileNotFoundError, ValueError, ValidationError, KeyError) as exc:
    engine_error = f"Could not resolve engine configuration: {exc}"

else:
    try:
        engine = create_cdm_engine(resolved)
        engine_created = True
    except RuntimeError as exc:
        engine_error = str(exc)

[gkennos]

this API change makes it technically breaking - shouldn't matter if this is being absorbed into the intended 2.0 upversion anyway?