validate: take the report's filename from core's `validate()` return value

[rbeezer]

Core's validate() now returns the report's filename, as of PreTeXtBook/pretext#3216 (merged 2026-09-15). It returns None when no report was written, which happens when method="server" and the validation server cannot be reached or reports an error. The parameters are unchanged, so nothing breaks today: Target.validate_source() calls it with keyword arguments and ignores the result.

Taking the filename from the return value would let two things come out of pretext/project/__init__.py:

• A second copy of core's naming rule. Line 921 rebuilds <source stem>-validation.txt. If core ever names the report differently, the CLI finds no file and reports "Validation could not be performed".
• An existence test standing in for "did core write a report?" (lines 922-924). Nothing removes an earlier report from logs/ first, so when core writes nothing, the report from a previous run is read as this run's result, and pretext validate reports that run's message counts.

Roughly:

        try:
            report_name = core.validate(
                ...
            )
        except OSError as e:
            log.debug(f"Validation could not run: {e}")
            return None
        finally:
            ...
        if report_name is None:
            # core writes no report when it cannot use the validation server
            return None
        report = Path(report_name)

What comes back is the path as a string: out_file exactly as passed, or else dest_dir joined with <source stem>-validation.txt. The CLI passes an absolute dest_dir, so it would receive an absolute path. A missing jing still raises OSError, which validate_source() already catches.

The earlier notes on core's validation rework are in #1189.

Claude Opus 5, acting as a coding assistant for Rob Beezer