fix(search): restore unquoted multi-word date keywords via pre-parse quoting

"added:previous month" returned HTTP 400 after the whoosh-compat
migration. The unquoted spelling was never parser-native anywhere: v2
rewrote it to explicit bracket ranges app-side before whoosh saw the
string, and the deleted translation layer consumed it itself, so users
and saved views have relied on it continuously while whoosh-compat
deliberately scopes it out of its parser (its DIVERGENCES.md entry 19)
and understands the phrases natively only as quoted values.

parse_user_query now quotes the closed six-phrase vocabulary (previous
week/month/quarter/year, this month/year) when it directly follows a
date field's colon, before parsing. Only quoting happens app-side; every
date computation stays in whoosh-compat's grammar, unlike v2's rewrite,
which computed the ranges itself. Date field names derive from
PUBLIC_FIELDS, the field name matches case-sensitively (the parser's own
field tagging is case-sensitive), the phrase case-insensitively (the
grammar accepts any case in the quoted form), and already-quoted
spellings, TEXT fields, unfielded words and bracketed ranges are
untouched.

The previously xfailed end-to-end regression test now passes as a plain
test, and a new acceptance class pins unquoted == quoted == mixed-case
result sets on a boundary fixture, no-error parsing for the whole
vocabulary across all three date fields, and that "title:previous month"
stays an ordinary text search. docs/usage.md now states the two
spellings are equivalent after a date field.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WMsn6DgzbvSqh1pwy66VVF

src/documents/search/_query.py

@@ -13,6 +13,7 @@ from whoosh_compat.errors import Diagnostic
 from whoosh_compat.errors import DiagnosticKind
 from whoosh_compat.errors import UnsupportedQueryError
 
+from documents.search._fields import PUBLIC_FIELDS
 from documents.search._registry import get_field_registry
 from documents.search._tokenizer import simple_search_tokens
 
@@ -70,6 +71,59 @@ _REGEX_TIMEOUT: Final[float] = 1.0
 # Uses Unicode properties to cover all blocks including Extension B+ planes.
 _CJK_RE: Final = regex.compile(r"[\p{Han}\p{Hiragana}\p{Katakana}\p{Hangul}]+")
 
+# The closed multi-word date-keyword vocabulary, unchanged since paperless
+# v2's rewrite_natural_date_keywords. whoosh-compat's date grammar
+# understands every one of these natively, but only as a QUOTED value
+# (its DIVERGENCES.md entry 19: unquoted multi-word values split at
+# whitespace, faithfully to whoosh); the unquoted spelling has been
+# honored continuously since the whoosh era by an app-level assist, so
+# _quote_date_keyword_phrases below keeps honoring it by inserting the
+# quotes and nothing else. Single-word keywords (today, yesterday) parse
+# unquoted already and need no entry.
+_DATE_KEYWORD_PHRASES: Final = (
+    "previous week",
+    "previous month",
+    "previous quarter",
+    "previous year",
+    "this month",
+    "this year",
+)
+
+# Field names are case-sensitive (matching the parser's own field
+# tagging); the keyword phrase is case-insensitive (matching the date
+# grammar's leniency for the quoted form). Date fields derived from
+# PUBLIC_FIELDS, never hand-listed.
+_DATE_KEYWORD_PHRASE_RE: Final = regex.compile(
+    r"\b("
+    + "|".join(
+        regex.escape(f.name)
+        for f in PUBLIC_FIELDS
+        if f.kind in (wc.FieldKind.DATE, wc.FieldKind.DATETIME)
+    )
+    + r"):((?i:"
+    + "|".join(_DATE_KEYWORD_PHRASES)
+    + r"))\b",
+)
+
+
+def _quote_date_keyword_phrases(raw_query: str) -> str:
+    """Quote unquoted multi-word date keyword phrases on date fields.
+
+    ``added:previous month`` becomes ``added:"previous month"``; the
+    already-quoted spellings don't match the pattern (the colon must be
+    followed directly by the phrase), and the same words after a TEXT
+    field or standing alone are ordinary text and untouched. Only quoting
+    happens here: every date computation stays in whoosh-compat's
+    grammar, which parses exactly this phrase vocabulary as quoted
+    values. This is deliberately NOT a revival of the deleted
+    translation layer, which computed the ranges app-side.
+    """
+    return _DATE_KEYWORD_PHRASE_RE.sub(
+        r'\1:"\2"',
+        raw_query,
+        timeout=_REGEX_TIMEOUT,
+    )
+
 
 def _has_cjk(text: str) -> bool:
     """Return True if text contains any CJK characters."""
@@ -285,7 +339,10 @@ def parse_user_query(
     """
     Parse user query through whoosh-compat, then blend in fuzzy/CJK clauses.
 
-    1. wc.parse() against the shared FieldRegistry (whoosh grammar -> AST).
+    1. Unquoted multi-word date keyword phrases on date fields are quoted
+       (_quote_date_keyword_phrases) so the historically honored
+       "added:previous month" spelling keeps working; then wc.parse()
+       against the shared FieldRegistry (whoosh grammar -> AST).
     2. Any diagnostics (bad dates/numbers) map to SearchQueryError subclasses
        and raise — the view returns HTTP 400 with every offending field
        listed, not just the first.
@@ -303,6 +360,7 @@ def parse_user_query(
        never went through the pre-whoosh-compat translation layer either.
     """
     registry = get_field_registry(settings.SEARCH_LANGUAGE)
+    raw_query = _quote_date_keyword_phrases(raw_query)
     result = wc.parse(
         raw_query,
         registry=registry,