BUG: Reuse a single Selection during table-format HDFStore.select reads

[Copilot]

Repeated reads from table-format HDF keys were constructing selection state twice per select call in the read path. This added unnecessary allocation/retention pressure and is consistent with the long-standing memory-growth behavior reported for repeated table reads.

• Read-path change

  • AppendableFrameTable.read now creates one Selection and reuses it for both:
    • axis extraction (_read_axes)
    • post-read filtering/reindexing (process_axes)
  • _read_axes was updated to accept an existing Selection instead of rebuilding from where/start/stop.

• API/internal contract cleanup

  • Table._read_axes signature changed from (where, start, stop) to (selection), making ownership of selection state explicit and avoiding duplicate construction.

• Regression coverage

  • Added a focused test in pandas/tests/io/pytables/test_select.py that monkeypatches Selection.__init__ and asserts only one initialization per temp_hdfstore.select("df"), while verifying result equality.

# before
result = self._read_axes(where=where, start=start, stop=stop)
selection = Selection(self, where=where, start=start, stop=stop)
df = self.process_axes(df, selection=selection, columns=columns)

# after
selection = Selection(self, where=where, start=start, stop=stop)
result = self._read_axes(selection=selection)
df = self.process_axes(df, selection=selection, columns=columns)

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.