feat: Font.color non-mutation + UTC-aware datetimes + Shapes.by_name (Modernization Phase 2)

[MHoroszowski]

Phase 2 of issue #29 (Modernization & Ergonomics): bug fixes + by_name

Phase 1 (PR #39) shipped PathLike + PERCENT_40 typo + Slide.background fix. This PR closes three nagging upstream tickets in a single bundle, all from issue #29's "bug fixes that pollute every other epic if left unfixed" group plus the most-cited shape-lookup ergonomic from the API ergonomics group.

Public surface — bug fixes

# Font.color is now non-mutating on read (closes scanny#1111, #1074)
font = run.font
_ = font.color.rgb            # returns None on unstyled run; ZERO XML diff
_ = font.color.type           # returns None on unstyled run; ZERO XML diff
font.color.rgb = RGBColor(...)  # still works — lazy <a:solidFill/> materialization

# core_properties datetimes are tz-aware on read (closes scanny#957)
prs.core_properties.created   # tz-aware datetime when XML carries 'Z' or ±HH:MM
prs.core_properties.created = dt.datetime(2024, 7, 4, 10, 0, 0, tzinfo=PDT)
# ↑ written as UTC: 2024-07-04T17:00:00Z

# Shape lookup helper (closes scanny#798, #309, #532)
slide.shapes.by_name("Title 1")  # → BaseShape  (first match, case-sensitive)
slide.shapes.by_name("Bogus")    # → KeyError

What it adds

Font.color non-mutation (closes scanny#1111, scanny#1074)

• Font.color returns a new _LazyFontColorFormat proxy that mirrors ColorFormat's public surface.
• Reads (type, rgb, theme_color, brightness, transparency) return None on a run with no <a:solidFill/> — the inherit signal — without writing to XML.
• Writes (rgb=, theme_color=, brightness=, transparency=) materialize <a:solidFill/> lazily on first set, then delegate to the real ColorFormat.
• A pre-commit Forge audit caught my initial draft returning MSO_THEME_COLOR.NOT_THEME_COLOR / 0.0 on the no-fill path — those are real settable values, not inherit signals. Conflating them would silently break inheritance on read/write round-trips. Fixed to return None for all four reads on the no-fill path.

UTC-aware datetimes (closes scanny#957)

• _parse_W3CDTF_to_datetime returns tz-aware datetimes when the source string carries timezone info: Z → UTC; ±HH:MM → fixed-offset datetime.timezone.
• Strings without offset markers (year-only, year-month, year-month-day, bare timestamp) continue to return naive datetimes — we don't assume a timezone where none was given.
• _set_element_datetime accepts both naive and tz-aware datetimes; tz-aware inputs are converted to UTC via astimezone(timezone.utc) before serialization, so the on-disk W3CDTF form always uses the canonical YYYY-MM-DDTHH:MM:SSZ shape.
• Sign-convention fix: prior _offset_dt had sign_factor = -1 if "+" else 1 (inverted) and adjusted clock values; new _tzinfo_from_offset_str uses POSIX-correct sign_factor = 1 if "+" else -1 and returns a tzinfo instance. Forge audit walked through '-08:00' and '+05:30' — same effective UTC instant produced in both old and new code (old returned naive-but-UTC-shifted, new returns tz-aware-with-original-offset).

_BaseShapes.by_name(name) (closes scanny#798, scanny#309, scanny#532)

• Returns the first shape in document order whose .name matches.
• Case-sensitive (matches PowerPoint's behavior).
• Raises KeyError with a clear message on miss.
• Defined on _BaseShapes so it inherits to SlideShapes, SlideLayoutShapes, SlideMasterShapes, etc.

Behavior changes (intentional, called out for downstream consumers)

• Reading font.color no longer inserts <a:solidFill/>. Code relying on the mutation as a side-effect must call font.fill.solid() explicitly.
• Reading core_properties.created / last_printed / modified on a document whose XML carries Z or ±HH:MM markers now returns a tz-aware datetime. Tests and callers that expected naive datetimes need updating; this PR updates the two existing test fixtures that exercised this path.

Tests

• 30 new pytest cases in tests/test_modernization_phase2.py covering byte-stable XML on Font.color reads, lazy materialization on first set, the inherit-signal contract (None for all four reads on no-fill), datetime parser tz-awareness across all five W3CDTF input forms, round-trip through save/reload, and the five by_name paths (hit, miss, case-sensitivity, multi-match, inheritance to layouts and masters). Two existing test fixtures updated to expect tz-aware values where the input strings carry Z. Full pytest: 3456 passed.
• 5 new behave scenarios in features/modernization-phase2.feature. Existing features/steps/coreprops.py updated to use tz-aware datetimes for the now-tz-aware reload values. Full behave: 1041 scenarios passed, 0 failed.
• Ruff clean: ruff check src tests → All checks passed; ruff format --check → no diff.

Reporting contract (CLAUDE.md §7)

$ python3 -m pytest tests/ -q | tail -3
........................................................................ [ 97%]
........................................................................ [100%]
3456 passed in 4.82s

$ python3 -m ruff check src tests | tail -3
All checks passed!

$ python3 -m behave features/ --no-color 2>&1 | tail -3
1041 scenarios passed, 0 failed, 0 skipped
3134 steps passed, 0 failed, 0 skipped
Took 0min 1.591s

UAT

• uat_modernization_phase2.py (untracked per CLAUDE.md §6) at repo root.
• Three demonstrations: byte-diff non-mutation, datetime round-trip with PDT input → UTC reload, by_name hit/miss.
• All programmatic assertions pass; visual UAT signed off by maintainer.

Skipped from issue #29 Phase 2 (deferred or no-op)

• collections.abc import path migration (would close Compaitibility script no longer breaks to collections changes in 3.10 scanny/python-pptx#771): already complete in this fork — every relevant import already reads from collections.abc import X. No-op.
• iter_leaf_shapes, Mapping ABC, find_by_xpath, selection-pane order listing: deferred to a future Phase 4 to keep Phase 2 focused on bug fixes.

Refs #29