docs(waterdata): correct include_hash_ids stability claims

The include_hash_ids docstrings claimed the dropped columns "are not
stable across record refreshes." That is true for the per-record version
UUID (daily_id/continuous_id/computation_id), which is regenerated on
every refresh, but wrong for the secondary hashes (time_series_id,
parent_time_series_id, field_series_id/field_measurements_series_id):
real-data checks confirm these are stable and are the documented join
keys back to the metadata endpoints (e.g. time_series_id links a values
row to get_time_series_metadata; the API docs recommend (time,
time_series_id) to identify an observation over time).

Rewrite the 11 OGC getter blocks, the _HASH_ID_COLUMNS comment, and the
get_stats_data block to distinguish the unstable per-record UUID from the
stable-but-opaque join keys, and tell callers to set include_hash_ids=True
(or name the column in properties) when they need to join, trace series
lineage, or disambiguate series sharing the same
(monitoring_location_id, parameter_code, statistic_id).

Docs only; no behavior change.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: thodson-usgs