Revisit the code that populates the search results table

[danielballan]

It works well sometimes, but other times it is comically slow, and it's hard to detect a pattern. I suggest we start by refactoring the worker to make it update a row at a time rather than a cell at a time and see how far that gets us.

bluesky-widgets/bluesky_widgets/qt/_search_results.py

Lines 23 to 32 in c2edf3c

	def _load_data(get_data, indexes):
	"Load a batch of data. This is run in a threadpool."
	for index in indexes:
	row, column = index.row(), index.column()
	try:
	item = get_data(row, column)
	except Exception:
	logger.exception("Error while loading search results")
	continue
	yield index, item

[cjtitus]

Part of the explanation for the slowness of cell-based updates is probably bluesky/tiled#645. Row-based updates will be faster, but still subject to really inefficient paging until that is fixed.

Once the issue is fixed, row-based updates will be ok, but still subject to latency on a per-row basis. I have code that does efficient chunk-based updates, which work well, though they will also get faster once bluesky/tiled#645 is fixed.