docs: add pcapkit.__version__ entry and the missing corekit.enum page

[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort)
• make test passes, and a test case covers the change
• Added a changelog entry -- N/A -- changelog centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

• docs — documentation only

---

Description of your pull request and other information

Closes #902.

• pcapkit.__version__: added under a new "Package Metadata" section on
  docs/source/pcapkit/index.rst, matching where other top-level pcapkit
  data lives -- it had no API entry, only prose mentions.
• pcapkit.corekit.enum had no doc page at all: added
  docs/source/pcapkit/corekit/enum.rst (shape matches module.rst/field.rst)
  and wired it into corekit/index.rst's toctree. EnumLookup/EnumRegistry
  now get real id anchors, and every :class:~pcapkit.corekit.enum.*``
  reference in contributing/conventions.rst resolves to a real `href` --
  verified against a built HTML tree, not a warning count, since `conf.py`
  sets no `nitpicky` and the build uses no `-W`.
• :file: audit: swept 498 occurrences (210 unique targets) across
  docs/source/, pcapkit/ and tests/. Fixed 7 wrong ones: two abbreviated
  pcapkit/protocols/{internet/hip,transport/tcp}.py citations missing their
  directory prefix in tests/protocols/test_option_roundtrip_unit.py, and
  five references in tests/corekit/test_sentinel_exports_unit.py still
  pointing at docs/source/conventions.rst's pre-refactor(docs): move the process pages out of the docs top level #901-move location. The rest
  are either external files, gitignored generated fixtures, deliberate
  historical references, or abbreviated mentions with directory context
  already established in the same file -- not defects.

Sphinx warnings: 58 on origin/main, 61 here (+3), all
duplicate object description for the three private methods explicitly
re-automethod'd in enum.rst -- the same pattern infoclass.rst already
uses for Info.__init_subclass__/__post_init__.

Docs-only change (plus two comment-only :file: fixes in existing tests, no
behaviour touched), so there is no coverage delta to claim.

[JarryShaw]

GOOD TO GO — cross-review verdict at head 55addee75, produced on Opus (author was Sonnet), briefed to falsify. Nothing load-bearing overturned.

Independently re-derived on the built tree, not from warning counts: the docs build exits 0 on both trees; the warning delta is exactly +3, all duplicate object description for the three explicitly-redeclared private methods; docs/source/pcapkit/corekit/infoclass.rst:11-19 is a real precedent that already contributes 2 identical warnings to the baseline, and its Info.__init_subclass__ renders twice with an id="id0" anchor on unmodified main — so this is pre-existing house behaviour, not something this PR introduces. id="pcapkit.__version__" resolves against pcapkit, not pcapkit.all.

Three corrections to the numbers, none blocking, all measured by me:

• :file: occurrences are 517, not 498. git grep -o -e ':file:[^]*' -- docs/source pcapkit testson3766c3c: 517 occurrences (413 in tests/, 74 in docs/source/, 30 in pcapkit/`), and 211 unique targets, not 210. Unchanged at 517 on this head, as expected — the fixes rewrite paths, they do not add roles. The classification work is unaffected; 15 of the ~72 non-existent targets were re-sampled and none is a genuine stale path.
• The five failures are not all one test. On unmodified 3766c3c09: three are subTests of SentinelExportTests.test_star_import_hands_back_the_canonical_object (module= pcapkit.corekit.module, pcapkit.corekit.fields.field, pcapkit.corekit.enum), and two are NoValueIsTheDocumentedFieldDefaultTests.test_an_undefaulted_field_reports_the_sentinel and test_setting_and_deleting_a_default_round_trips_through_the_sentinel. Same root cause, so "pre-existing isolation hazard" stands, but the attribution in the hand-back over-stated it.
• Relayed unverified: the cross-references into enum.html were counted at 24 hrefs over 9 unique targets against a baseline of 0. I did not rebuild to re-derive that; the direction is what matters and it agrees.

CI at this head: 53 CheckRuns pass, 0 fail, the five Python 3.10–3.14 legs still in progress. Unpublished and unmerged — yours to merge.