docs: restructure usage examples and add context7 configuration

[paullla]

Q	A
Bug fix?	no
New feature?	no
BC breaks?	no
Deprecations?	no
Issues	-
License	MIT

Problem

The README's usage examples sit under a single vague "Usage" heading as bold pseudo-headings, which reads fine on GitHub but retrieves poorly in documentation indexes such as Context7 that serve snippets to AI coding agents by nearest heading. PATCH is the recommended way to edit translations but the only edit example showed the PUT payload with ids, there was no response example for ?groups[]=translations, and known limitations (filtering by translated fields, the per-item translation query on collections) were undocumented, which leads agents to invent answers instead of stating they are not supported. The bundle is also not yet indexed on Context7, so agents answer questions about it from stale training data.

Changes

• Restructure the usage section into task-oriented headings, each with a self-contained example: request a single locale (query param and Accept-Language as full request lines), return all translations (now with a response example showing the group is additive), create with POST, update with PATCH, replace with PUT.
• Add a merge patch example: one locale submitted, no ids needed, other locales untouched.
• Prefix the PUT example with its full-replace semantics and a link to the standard_put notes.
• Add a limitations section: no filtering/ordering by translated fields, and one translation query per collection item (EXTRA_LAZY, no built-in eager loading).
• Normalize code fence info strings (``` php to ```php).
• Add context7.json, validated against the published schema: indexes only README, changelog and upgrade guides (excludes src, tests, config, bin), eleven usage rules covering the conventions and footguns (serialization groups, payload shape, PATCH over PUT, standard_put, locale selection), versioned docs for the v2.0.1 and v1.4.1 tags, and the ownership verification keys for the Context7 entry.