Windows: perry-runtime --lib test suite has pre-existing failures now that the crate compiles there (#7355 fallout inventory)

[proggeramlug]

#7355 made perry-runtime compile on Windows for the first time (main's unguarded eh_walker calls had broken the build outright, and no Windows CI existed to notice). Running cargo test -p perry-runtime --lib on a real Windows 11 host now surfaces pre-existing platform gaps that were previously unmeasurable. None are related to the #7354 walker — gc::roots::stack_maps and exception:: tests are green, and cargo test -p perry-codegen --lib is 599/599.

Inventory from sequential runs (--test-threads=1, iteratively skipping each process-killing stopper; ~904 tests pass between stoppers):

Process-killing stoppers (each aborts the whole harness, hiding everything after it):

• dyn_eval:: throw-exercising tests — js_throw raises SEH; the unwind dies with "panic in a function that cannot unwind" / STATUS_STACK_BUFFER_OVERRUN. Backtrace shows RtlUnwind → _report_gsfailure from perry_runtime::exception::js_throw under dyn_eval::tests::catch_throw. The Windows exception transport does not survive under the Rust test harness (GS cookie failure during unwind through Rust frames).
• ffi::setjmp::tests::setjmp_smoke_via_c_int_buffer — STATUS_ACCESS_VIOLATION (setjmp is already known host-gated on Windows; issue_856_promise_microtask_setjmp_does_not_crash passes).
• native_abi::tests::buffer_span_guards_require_registered_buffer — non-unwinding panic → abort.

Plain assertion failures:

• date::tests::utc_getters_ignore_process_timezone — date.rs:1720, UTC getters are not ignoring the process timezone on Windows (host TZ was UTC+2 DST).
• gc::tests::incremental_sweep_reclaim::budgeted_reclaim_runs_process_malloc_trim — asserts allocator trim ran, but telemetry reports malloc_trim: {reason: "not_supported", status: "unsupported"} on Windows; the [gc] Graduate incremental collection: budgeted scanner coverage, assists for all phases, default-on (successor to #6083) #6180 RSS-floor assert assumes a glibc/mimalloc trim.
• gc::tests::teardown::map_set_side_allocations_release_on_thread_exit — teardown.rs:23, side-allocation count 66 != 64 at thread exit.
• child_process::tests::test_spawn_sync_result_fields — failed in a parallel run (not yet re-isolated).

More failures may hide behind the stoppers — each abort truncates the run, so this list is a lower bound. Fixing the SEH-transport-under-test-harness stopper first would let the rest be inventoried in one pass.