feat(text): Field authoring — _Paragraph.add_field, _Field class, CT_TextField.text setter (Phase 3)

[MHoroszowski]

Phase 3 of #20 — Field authoring API

Ports scanny/python-pptx#797 ("Added a:fld type to paragraphs for page numbers and datetimes") onto Phase 1's (PR #48) OOXML primitives and Phase 2's (PR #49) public slide/master surface. Manually re-derived against this fork's ruff-formatted, post-Phase-1 source per CLAUDE.md §2 — git cherry-pick would conflict on every touched file because upstream predates the repo-wide ruff format pass (PR #10).

What this PR lets users do

p = slide.placeholders[1].text_frame.paragraphs[0]
p.clear()
fld = p.add_field()
fld.type = "slidenum"     # or "datetime1".."datetime13", "title", etc.
fld.text = "<#>"          # the placeholder glyph PowerPoint shows before resolving
prs.save("out.pptx")

Open the resulting .pptx in PowerPoint or Keynote and the slide number renders automatically. The id GUID is generated transparently in the format PowerPoint emits when the user runs Insert → Slide Number.

Changes

• pptx.oxml.simpletypes.ST_FieldType (NEW) — XsdString subclass for the a:fld@type attribute value. Phase 1 used a plain XsdString placeholder; this is the named subtype.
• pptx.oxml.text.CT_TextField.text — read-only property from Phase 1 now has a setter. Writes through get_or_add_t() and routes the value through CT_TextField._escape_ctrl_chars (NEW static method) which replaces chars in [\x00-\x08\x0B-\x1F] with _xNNNN_ uppercase-hex form per OOXML §22.9.2.19, leaving \t (0x09) and \n (0x0A) alone.
• pptx.oxml.text.CT_TextParagraph.fld — ZeroOrMore("a:fld", successors=("a:endParaRPr",)) accessor. The a:pPr successor tuple already named a:fld per Phase 1 (forward declaration). xmlchemy auto-generates _add_fld() from the ZeroOrMore.
• pptx.text.text._Field (NEW) — public-via-add_field-return-value class wrapping <a:fld>. Leading-underscore private name matches _Run and _Paragraph. Properties: font (Font wrapping rPr), text (read/write, routes through the escaping setter), type (read/write, str | None).
• pptx.text.text._Paragraph.add_field() (NEW) — appends a fresh <a:fld> with a uuid4 GUID id wrapped in braces, uppercase hex. Returns a _Field; caller sets type and optionally text. Run-style symmetry is deliberate — users who know add_run() don't have to learn a new pattern.

Out of scope (deliberate)

• Field discovery during paragraph iteration — p.runs continues to yield only _Run objects. Phase 4 will surface _Field instances via a separate accessor.
• HandoutMaster Python class and watermark helper — Phase 5.
• MSO_FIELD_TYPE enum — type stays plain str for now to mirror Added a:fld type to paragraphs for page numbers and datetimes scanny/python-pptx#797. An enum can land in a later cleanup once the canonical field-type list is settled.

Verification (local, CPython 3.14.4)

python3 -m pytest tests/ -q                       → 3626 passed in 5.32s (+28 vs Phase 2 baseline)
python3 -m ruff check src tests                   → All checks passed!
python3 -m ruff format --check src tests          → 216 files already formatted
python3 -m behave features/ --no-color            → 1048 scenarios, 0 failed
python3 uat/uat_headers_footers_phase3.py         → PASS (id={2ED44585-07B5-4BC8-93B2-49122D50BCC2}
                                                          + type='slidenum' + text='<#>' all round-tripped)

28 new tests:

• 14 in tests/oxml/test_text.py — CT_TextField setter, _escape_ctrl_chars (BEL/tab/newline edges), CT_TextParagraph.fld/_add_fld.
• 14 in tests/text/test_text.py — Describe_Field (10 tests: font, text get/set with escape, type get/set/clear) + Describe_Paragraph_add_field (4 tests: GUID regex match, distinct ids on consecutive calls, fld-after-runs ordering, parent chain).

Refs #20.