fix: gate shorebird_updater_supported on having a known rust target

[eseidel]

Summary

build_rust_updater.gni previously set shorebird_updater_supported from target_os alone (is_android || is_ios || ...). But GN evaluates BUILD files under every toolchain in use, including host and sub toolchains that can hit (target_os, target_cpu) combos we don't explicitly handle. When such a toolchain entered the if (shorebird_updater_supported) block but none of the per-cpu cases set shorebird_updater_rust_target, GN blew up:

ERROR at build_rust_updater.gni:72: Undefined identifier in string expansion.
  shorebird_updater_output_lib = "$...$shorebird_updater_rust_target..."
"shorebird_updater_rust_target" is not currently in scope.

Reproduced on Windows host builds of android engine variants: win_build.sh's gn --android --android-cpu=arm64 invocation, evaluated under the Windows host toolchain, landed on an is_win/target_cpu combo our is_win branch doesn't cover.

Fix

Flip the logic. Compute shorebird_updater_rust_target first (default empty), then derive shorebird_updater_supported from whether it got assigned. Unhandled combos become unsupported, so consuming BUILD.gn files' existing if (shorebird_updater_supported) gate cleanly skips the updater for that toolchain — which is correct, since the updater only needs to be built for toolchains whose final binary is the engine.

No new os/cpu combos added. No behavior change for already-handled combos. Just safer unhandled-combo behavior.

Tracking: shorebirdtech/_shorebird#2014
Follow-up to: #124, #127, #128

[bdero]

Clearly, my original change to add support for building the Rust updater in GN was not complete. I'd consider just reverting the whole thing to unblock things if continues to give you trouble.

[eseidel]

No no, this is very good. This is taking us in the right direction. I'm glad for us to finally do this.

[eseidel]

The Clankers are extremely good at this.

[eseidel]

I asked Claude if we should move to the full "just call the Rust compiler directly" solution. It gave me this long history of how Fuchsia and Chromium had started with Cargo like we are, and then slowly migrated to calling the Rust compiler directly. It recommended that we stay where we were, just calling Cargo for now, even though it added all this extra noise, because the full solution was a lot more work.