docs: use the glossary's terms for collections and locations

[thodson-usgs]

Why

A glossary pass over the package, scoped to the modules that are not
deprecated. Prose says collection where it meant one of a service's named
sets of records, and monitoring location where it meant a place measurements
are recorded.

Parameters are untouched throughout. site_type=, site_type_code= and the
sites_df example variable are the services' own vocabulary and stay as they
are — that is the anti-corruption boundary working, not drift.

What changed

• waterdata/ — measurements, metadata, ratings, samples,
  time_series: docstring prose and example comments.
• ogc/planning.py — shared OGC machinery serves both Water Data and NGWMN, so
  nothing in it is about one service's spelling.
• wqp.py — _query_wqp's docstring and the module todo. WQP's profiles
  (Result, Station, Activity) are collections; the reference to
  services_wqx3 now notes that the name itself is legacy.
• CONTEXT.md — one Known legacy names bullet, below.

Legacy NWIS is deliberately excluded

service= on the NWIS getters does name a collection. It is also frozen public
surface on a module being retired, so the parameter keeps its name and its
docstrings keep the same word — describing a parameter in a term the module
itself never uses helps nobody, and the module is not where new vocabulary
should be landing. The CONTEXT.md bullet records that scope.

NEWS.md is left alone for the same reason. The two entries this originally
touched are both about deprecated NWIS, and shipped release notes are a
historical record rather than prose to bring into line.

Prose only

No executable code changed, verified by comparing the AST of every touched
module against upstream/main with docstrings stripped:

7 .py files changed; executable code differs in: NONE

1139 tests pass, ruff format --check clean.

Deferred, with a home

• wqp.services_wqx3 / wqp.services_legacy hold collection names and are
  public. Tracked in Glossary violations across the package: full inventory for planning #406 §3 — it needs an ADR 0012 deprecation cycle, not a
  prose PR.
• The censored → partly dated wording in nwis.py, its tests, and the
  08/27 NEWS entry. A genuine terminology problem (in water data a censored
  value is one outside a detection limit — a meaning this package ships in
  wqp.what_detection_limits), but it is confined to deprecated NWIS and its
  release notes, so it is out of scope here. Worth its own small PR if wanted.
• The pandas UserWarning raised in the first draft of this PR is fixed in
  fix(nwis): name the peak date format instead of making pandas guess #403.

Relationship to the other open PRs

Based directly on main. Touches no file #396 or #400 touches except
CONTEXT.md.

#405 overlaps here. It restructures Known legacy names and, under ADR
0013, moves domain terms at an adapter's surface out of that section and into
the term's own entry — where NWIS's service= is already named. Whichever
merges second should fold this bullet into the Collection entry rather than
keep both.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JAEQqs7XzQHGQQi2KakuXD