[ItIsUday]

Why

The robot runner's elapsed/parallelization numbers can't be computed because robot test cases land with no real start time — they fall back to ingest time (all cases in a run cluster within a few milliseconds). Root cause is in this plugin: Robot's output.xml records a per-suite starttime, but parse_func read it only to derive each case's duration and then discarded it, emitting a single flat <testsuite name="robot"> with no timestamp. Downstream, the suite timestamp is what the server resolves and stamps onto every case in the suite, so with none emitted, robot timing was lost.

What changed

• Emit one <testsuite> per leaf suite, each tagged with its own start timestamp, instead of one flat robot suite — so each case inherits its real suite start.
• Normalize the timestamp to ISO-8601 naive (no timezone): RF < 7.0's starttime (20210319 11:34:50.085) is reshaped via strptime(...).isoformat(); RF 7.0+'s already-ISO start passes through untouched. Timezone resolution stays server-side.
• Absent/malformed start times degrade to None (server falls back as before) rather than failing the upload.
• Test paths are unchanged (built from each case's classname, not the suite element), so existing record/subset goldens still match. Added a parse_func test asserting per-suite timestamps.