feat(symcache): Build symcaches for WASM modules from Emscripten source maps

[d2anamaria]

Summary

Emscripten can ship a Source Map v3 file next to a .wasm module instead of DWARF. This change
turns such a module plus its map into a SymCache, so WASM builds without DWARF become
symbolicatable. Everything sits behind the wasm-sourcemap feature flag, which keeps the
sourcemap dependency out of the way for consumers that don't need it.

Nothing calls this yet. The symbolicator wiring is a follow-up.

Why a source map works as debug info

An Emscripten WASM map abuses the source map format:

Source map field	Normal JS meaning	Emscripten WASM meaning
destination line	line in the minified file	always 0
destination column	column in the minified file	byte offset into the module
source + source line	original location	original location

Why SymCache

That byte offset is already a symcache key. SourceMapCache is keyed by a line and a column
instead, so using it would mean faking a position and unwinding it on every lookup. Reusing SymCache
means WASM frames flow through the existing address lookup path, with no new format and no new
lookup API.

Inputs

• The .wasm module is required, not optional. A map has no notion of a function.
• Function bounds and names come from the module; lines come from the map.
• Both are joined by address.

What lands in the cache

• Every function body, even one the map says nothing about. Partial info is still useful, and it
  stops a lookup from bleeding into the previous function.
• Unnamed bodies get wasm-function[<index>], matching what browsers show.
• A range with no known location resolves to the function with no file and no line, instead of
  inheriting the location before it.
• Mappings outside any function body are dropped. They fall in the padding between bodies.
• An Emscripten one-field segment marks a hole and closes the preceding line range.

WASM object changes

Exact function body bounds and WASM function indices are now available. The existing symbol
iteration stretches each body to be contiguous with the next, hiding the padding between bodies.
Joining a map to stretched bounds would let padding inherit a line from the function before it.

Debug ids are out of scope

This change derives no identifier. It reads WasmObject::debug_id() as-is.

An earlier revision also made the build_id custom section length-prefix aware, per the
tool convention. That is
correct in isolation but breaks matching: @sentry/wasm reads the same section without stripping
the prefix, so a corrected reader here would file uploads under an id no event carries. The parsing
change was removed and the whole question moved to
#1069, which needs a coordinated decision
across symbolic, the SDK, and the CLI.

Whichever way that lands, this change needs no rework — the embedded id follows debug_id()
automatically.

Limitations

• Only plain maps. Index maps and Hermes maps are rejected.
• A map referencing an undeclared name is rejected.

[github-actions]

	Fails
🚫	Please consider adding a changelog entry for the next release.

	Messages
📖	Do not forget to update Sentry-docs with your feature once the pull request gets approved.

Instructions and example for changelog

Please add an entry to CHANGELOG.md to the "Unreleased" section. Make sure the entry includes this PR's number.

Example:

## Unreleased

### Features

- Build symcaches for WASM modules from Emscripten source maps ([#1070](https://github.com/getsentry/symbolic/pull/1070))

If none of the above apply, you can opt out of this check by adding #skip-changelog to the PR description or adding a skip-changelog label.

Generated by 🚫 dangerJS against aeebbc9

[loewenheim]

Once #1081 is merged the logic for preserving exact function sizes will be unnecessary (it will just work that way in general).