check notation uniqueness

[joergklausen]

feat:The checker uses the existing tables_en/wmdr-tables.csv mapping to discover the published WMDR registers, rather than duplicating that mapping in Python. That file currently maps the repository table IDs to the corresponding codes.wmo.int/wmdr/... registers.

It then:

• checks every local CSV with a notation column for duplicates within that file;
• checks every published register using its plain CSV endpoint with status=valid;
• checks observed-variable notations jointly across 1-01-01 … 1-01-05;
  checks observing-method notations jointly across 5-02-01, 5-02-03, 5-02-05;
• treats notation comparisons as case-sensitive, trims surrounding whitespace, and normalizes Unicode to NFC;
• fails with exit status 1 on duplicates, malformed tables, missing notations, or inability to check the registry;
• uses only the Python standard library.
  The special cross-table checks correspond directly to the repository's current division into five observed-variable tables and three observing-method tables.
  The workflow runs on every push and every pull_request, with a manual workflow_dispatch option as well. It uses the currently documented actions/checkout@v7 and actions/setup-python@v7 with Python 3.13.
  No requirements.txt addition is necessary.
  A successful run is deliberately compact:

WMDR notation uniqueness report

Repository CSVs: 65 table(s), ... notation(s)
Repository CSVs / within-table: OK - all notations are unique within each table
Repository CSVs / cross-table: OK - observed-variable and observing-method notations are unique across their related tables

Published registry: checking 65 register(s) at codes.wmo.int ...
Published registry: 65 table(s), ... notation(s)
Published registry / within-table: OK - all notations are unique within each table
Published registry / cross-table: OK - observed-variable and observing-method notations are unique across their related tables

RESULT: PASS - all checked notations are unique

Duplicates get their notation and locations reported, for example:

Repository CSVs / cross-table: FAIL - duplicate notation(s) across related tables
[observed variables] '123' occurs in 1-01-01, 1-01-03

Currently published atmosphere and terrestrial observing-method registers both contain inapplicable and unknown. Therefore these are currently legitimate cross-table duplicate findings. Flagging them is preferable initially.