[epifanio]

Refs #121.

Problem

The ECQL grammar only matches keywords in upper case, so these fail with UnexpectedToken:

from pygeofilter.parsers.ecql import parse
parse("attr like 'x%'")
parse("a = 1 and b = 2")
parse("intersects(geometry, POINT(1 1))")

ECQL keywords are case-insensitive, and the CQL2-text grammar here already matches them that way ("AND"i, …). As noted in #121, the ECQL grammar was left case-sensitive. That affects servers that parse user CQL with the ECQL parser. For example, pycsw parses CSW CQL_TEXT constraints with it for repositories that don't use a database. Long-standing CSW clients send apiso:Type like 'series', which pycsw's own cql2fes path accepts but this parser rejects.

Change

• ecql/grammar.lark: every keyword literal is now case-insensitive ("LIKE" → "LIKE"i, etc.), including the spatial and temporal predicates, ENVELOPE, and DOES-NOT-EXIST. Distance units and the shared WKT grammar are unchanged.
• ecql/parser.py: upper-case the spatial predicate name before the SPATIAL_PREDICATES_MAP lookup and the DWITHIN check. Those rules keep the matched token, so intersects(...) would otherwise raise a KeyError. This matches what the CQL2-text parser does with op.lower().

Tests

• test_keywords_case_insensitive: each keyword form (boolean, comparison, LIKE/ILIKE/IN/IS NULL, EXISTS/DOES-NOT-EXIST, INCLUDE/EXCLUDE, temporal, spatial, RELATE, DWITHIN/BEYOND, BBOX, ENVELOPE), in lower case and title case, must parse to the same AST as upper case. On main these 54 cases fail; with this change they pass.
• test_attribute_names_starting_with_keywords: notes, index, orbit, likes, isbn, android and between_x still lex as attribute names. These pass both before and after the change.
• tests/parsers: 369 passed. ruff check and ruff format --check with the pinned v0.11.13 are clean.