Document SUMMARY indexes for end users

SUMMARY indexes are now built automatically when a CTable is first closed, and they accelerate where() queries by skipping blocks that cannot match the predicate. This is a significant feature but it is only described in the API reference, not in any user-facing guide. The tutorial 15.indexing-ctables.ipynb covers FULL/PARTIAL indexes but does not mention SUMMARY at all.

Suggested work:
- Add a section to 15.indexing-ctables.ipynb explaining: what SUMMARY indexes store (per-block min/max), when they are built automatically, the create_summary_index=False opt-out, and how to verify they are being used.
- Add a short example that measures query time with and without a SUMMARY index on a realistic dataset
- Update doc/reference/ctable.rst to cross-reference the new create_summary_index parameter

Start with tutorial 15.indexing-ctables.ipynb and the SUMMARY index behavior described in the issue, then inspect doc/reference/ctable.rst for the create_summary_index parameter. Add the user-facing explanation, opt-out and verification details, plus a realistic timing example comparing queries with and without SUMMARY indexes. Done means both documentation locations cover the feature and cross-reference the parameter.

Written by the indexing model from the issue text.