standards write speed, UoA collection and metadata fixes

[bvkay]

Depends on IAGA-DVI-DataStandards/mt-io#6: item 2 imports UoACollection from there, so CI against the released mt-io will fail on that import until #6 merges. The other items are independent of it.

1. Bulk add_rows and one standards summary: the ~600-row standards summary table was written one h5py setitem per row, twice per file (file init and add_survey) and add_row formatted every row into a debug f-string even with logging off. Rows now write as one block and the summary array is built once per process and reused.

2. Split PR6-24 stations into runs: the UoA client now builds its file list through UoACollection (added in mt-io#6), so multi-run EDL stations come in as separate runs instead of one concatenated block.

3. Scan for data in blocks: has_data ran np.nonzero over entire channel datasets, so closing a file read every channel in full. It now scans ~4M-sample blocks and exits at the first nonzero.

4. Client metadata fixes: in phoenix.py the station update_metadata() call sat outside the station loop so only the last station was finalised and in metronix.py the survey update_metadata was referenced without parentheses.

[kujaku11]

@bvkay Thanks for the updates, these look like updates for efficiency.