test(base58): run the vector gate on Kotlin/Native, not just the JVM

[bmc08gt]

What

:libs:encryption:base58's Kotlin/Native vector test never compiled, so Base58VectorTest only ever ran on the JVM host despite the "asserted on both Android and iOS" claim in its own KDoc. This makes it genuinely run on both toolchains.

Why

The iosTest actual for readTestResource called NSString.stringWithContentsOfFile without opting in to @kotlinx.cinterop.ExperimentalForeignApi — a hard compile error on Kotlin/Native:

e: .../src/iosTest/kotlin/com/getcode/vendor/TestResources.ios.kt:15:18 This declaration needs opt-in.
   Its usage must be marked with '@kotlinx.cinterop.ExperimentalForeignApi'...
> Task :libs:encryption:base58:compileTestKotlinIosSimulatorArm64 FAILED

Adding the opt-in alone would only have moved the failure to runtime: Kotlin/Native test binaries ship no resource bundle, so NSBundle.pathForResource returns null.

This is not a parity coverage hole. Per test-vectors/README.md, base58's iOS parity is covered by the iOS repo's FlipcashCoreVectors via xcodebuild, and that stays green. What this closes is misleading dead source — and it makes the run matrix's parity claim true on both toolchains inside this repo rather than only documented.

How

Adopt the fix :libs:codes:kikcode already uses (#1287): a generateTestFixtures task that compiles src/commonTest/resources into commonTest as Kotlin constants. That removes the need for the expect/actual readTestResource entirely and works on every target — so all four TestResources*.kt files are deleted, including the JVM actuals that are now redundant.

Second instance of the same defect, found while verifying

Base58Test.kt sat in src/androidTest/. Under the com.android.kotlin.multiplatform.library plugin the host-test source set is androidHostTest — this module declares only withHostTest {}, and :libs:encryption:base58:tasks --all confirms no device-test task exists. So src/androidTest/ was not a source set and those 13 unit tests had never run either. Moved to androidHostTest.

Results

Read from the JUnit XML, not just the build exit code:

Target	Tests	Result
testAndroidHostTest	14 (1 vector + 13 unit)	green
iosSimulatorArm64Test	1 vector	green

The module's base58.json is still byte-identical to canonical (9566944…c8b19d).

Also

test-vectors/README.md: run-matrix row for base58 now names both toolchains; added a note under the base58 section explaining why there is no resource loader (so the next person doesn't reintroduce one); and corrected the Regenerate sync path, which said src/test/resources/ when the file actually lives in src/commonTest/resources/.

Reviewer note

The src/androidTest/-isn't-a-source-set trap applies to any module on the KMP Android plugin, and it fails silently — no error, the tests just never run. Worth a sweep if more modules have been converted to KMP since.