fix(document-processors): correct cascade escalation metric + review nits

Address round-1 review on PR #914:
- Attribute the OCR hop in a fast->structured->ocr inline cascade to
  from_tier="structured" (not a second "fast" escalation), so
  astrolabe_document_escalation_total per-tier counts stay accurate.
- Add test_inline_fast_structured_ocr_cascade pinning that two-hop path and the
  metric attribution.
- Note in classify_from_text that its doc-level control ratio is over full_text
  (all pages), not the sampled subset classify_pdf uses.
- Clarify that corrupt_glyphs never lands in the suppressed-escalation counter.
- Dedupe the glyph-corrupt test string into tests/fixtures/glyph_corruption.py.
- Use pytest.approx for the control-char-ratio zero checks (SonarCloud S1244).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

tests/fixtures/glyph_corruption.py

@@ -0,0 +1,11 @@
+"""Shared test data for the tier-0 glyph-corruption signal.
+
+A fast-tier text layer that looks like words -- normal spacing and token lengths,
+so it scores HIGH on ``_text_quality`` -- but leaks C0 control characters: the
+broken-/ToUnicode signature that ``classifier._control_char_ratio`` catches. The
+alphabetic tokens decode to a pangram under a -3 (Caesar) shift.
+
+Kept in one place so the classifier and registry tiering tests can't diverge.
+"""
+
+GLYPH_CORRUPT_TEXT = "WKH \x0f TXLFN \x10 EURZQ \x11 IRA MXPSV \x0f RYHU \x10 GRJ " * 6