feat: read v-prefixed tags and keep their prefix on the next tag

[lesnik512]

Closes #58.

semvertag now reads lowercase-v SemVer tags (v1.2.0) as bump baselines alongside bare ones, and the new tag inherits the latest tag's prefix: after v1.2.0 comes v1.3.0, after 1.2.0 comes 1.3.0. A repo whose history is entirely v-prefixed no longer reports no_tags forever.

Decisions (from the triage grilling on #58):

• Inherit, don't configure. The prefix comes from the latest tag, so no setting is needed and each repo keeps the convention it already uses. Reading v but always emitting bare would fix no_tags and break every v*-triggered release workflow instead. A force-prefix setting can be added if a real need shows up.
• Lowercase v only. V1.2.0, release-1.2.0 and the floating v0 are still not candidates. v0 never parses, so this repo keeps emitting bare release tags and just publish is unaffected.
• Precedence first, v wins exact ties. The highest version wins regardless of prefix. A tie such as 1.2.0 and v1.2.0 almost always means a repo re-tagged while moving to v, so v is the direction of travel.
• Output contract unchanged. schema_version stays "1.0" and no status token changes. The tag field and the Action's tag output can now carry v1.3.0; only their descriptions change.
• ADR-0003 amended in place, no new ADR. Its v sentence recorded a deferral and would contradict the code. The PEP 440 rejection beside it still stands. CONTEXT.md updates Latest tag, adds Tag prefix (bare "prefix" belongs to the branch-prefix strategy), and narrows Release tag to this repo's own convention.

Upgrade note (behaviour change): a repo that followed the old docs and seeded a bare 0.1.0 next to a v history (v1.4.0, ...) now bumps from the higher v tag and jumps to v1.5.0. Both provider guides call this out next to the seeding guidance.

Also updated: the no_tags reason and Rich message (seed 0.1.0, or v0.1.0 for v tags), README, docs/index.md, both provider guides, and the tag output description in action.yml and the GitHub guide.

Out of scope: a no_tags diagnostic for histories where tags exist but none parse (filed separately).

Tests go through SemvertagUseCase.__call__ with the stub provider: v inheritance, mixed history, the tie in both input orders, a v prerelease finalizing (v1.0.0-rc.1 + patch to v1.0.0), unrecognized tag forms, and AlreadyTagged on a v tag. With v stripping removed, every new test involving a v tag fails.

Verified with just lint-ci, just test-ci (493 passed, 100% coverage), just adr-check and just docs-build.