Deprecate query_engine.batch_size in favour of max_selectors_per_batch

Problem: The config field query_engine.batch_size (config_parser.py:401, default 500) is opaque. It actually caps the number of unique TimeSeriesSelector instances per solve batch (see build_batches, report_utils.py:51), not a generic row/record count.

This gets worse once PR #98 lands, which adds query_engine.max_containers_per_batch. Two batching dimensions sitting side by side, one named batch_size and one named max_containers_per_batch, is confusing. They should read symmetrically: max_selectors_per_batch and max_containers_per_batch.

Proposal: Rename batch_size to max_selectors_per_batch, keeping batch_size as a deprecated alias for one release.

• config_parser.py: add max_selectors_per_batch (default 500); accept batch_size as a deprecated alias that resolves to it and emits a DeprecationWarning; error if both are set.
• Update internal call sites and param names: report.py:913, report.py:936; the batch_size parameter in report_utils.build_batches / the batch helpers; docstring in calculated_channel.py:51.
• Update tests (report_utils_test.py, config_parser_test.py) and docs.

Done when: configs use max_selectors_per_batch; batch_size still works but warns; deprecation window and removal release are noted in the changelog/docs.

Start with config_parser.py around line 401 and trace build_batches in report_utils.py, then inspect the call sites in report.py and the documentation in calculated_channel.py. Run report_utils_test.py and config_parser_test.py before changing behavior. Done means the new name is used internally, the old alias warns and conflicts are rejected, and the deprecation window is documented.