GeezerVison

fix: keep the query on relative and root-absolute imports from served…

… modules (#2038)

ResolveSpecifierToPath dropped `?query#fragment` from every non-http
  specifier before consulting the import map or the referrer. A module
  served over HTTP that imports `/ns/asm?path=%2Fsrc%2FHome.vue` therefore
  resolved to `http://host/ns/asm`, which the Vite dev server answers with
  400 because the query is the module's identity. Any root-relative or
  relative `/ns/...` specifier carrying a query hit the same wall
  (`?path=`, `&mode=inline`, `?vue&type=`, `?ns_worker=1`), so every
  framework on the Vite dev flow was exposed; Vue merely hit it first.

  The seam now strips the query only once every HTTP outcome has
  returned, which is the iOS runtime's ordering: import-map lookup and
  HTTP-referrer resolution see the full specifier, and only filesystem
  probing sees the bare path. import() hands its specifier to the seam
  verbatim and routes on the resolved URL when the seam makes a relative
  or root-absolute spec HTTP, so those imports stay on the async graph
  walk instead of the blocking fallback.

feat: add DOMException and CustomEvent as lazy globals (#2027)

feat: add DOMException and CustomEvent as lazy globals (#2027)

feat: add TextEncoder/TextDecoder and atob/btoa on a lazy-global tier (…

…#2026)

fix: lock the remaining process-global state shared across isolates

The caches #2020 named (Console timers, ArgConverter, MetadataNode /
MetadataReader) were already covered by #2013: Console and ArgConverter moved to
per-runtime RuntimeState, MetadataNode got s_nodeCacheMutex, MetadataReader its
own StateMutex, and JEnv / MethodCache a shared_mutex each. Auditing the runtime
for what actually remained in that class turned up four more.

File::Buffer was a single process-wide 1MB scratch buffer that ReadText filled
and returned a pointer into. Main and worker runtimes read modules concurrently
from their own threads, so one thread's fread overwrote bytes another was still
copying out - silent module-source corruption rather than a crash, which is why
it never showed up in a tombstone. The buffer also never saved the allocation it
appears to save: every caller goes through the std::string overload, which
copies out of it on the next line. Reads now go straight into the returned
string, and the borrowing overload, which had no callers, is gone.

JType::EnsureInstance published *instance before Init had filled clazz, ctor and
valueMethodId, so a second thread could find the pointer non-null and call
through uninitialised JNI ids - on the hot boxing path.

CallbackHandlers::Init runs once per runtime from PrepareV8Runtime, so every
worker start rewrote the process-global class and method-id statics, and re-ran
MethodCache::Init, while the isolates already running were reading them.

MetadataNode::IsJavascriptKeyword filled a function-local static set behind an
empty() check, racing reads from every runtime's thread.

Also adds the missing ftell < 0 guard on the read path.

Verified: arm64-v8a native build, full test suite 1038/1038.

ci: unit test fixes

ci: unit test fixes

Merge remote-tracking branch 'origin/main' into feat/hmr-dev-sessions…

…-integrated

release: 9.0.5