Take a crashed file's cause only from what the run wrote after the last test, give a subtest's panic to that subtest, and keep the SDK's test count (issue 1901, review round 1)

go/buildstep/conformance/captured/node/logcrash.console

@@ -0,0 +1,2 @@
+npm notice run test
+npm notice run node --test --experimental-strip-types 'test/*.test.ts'

go/buildstep/conformance/captured/node/logcrash.tap

@@ -0,0 +1,42 @@
+TAP version 13
+# retrying after Error: the store was not ready yet
+# Subtest: passes and logs
+ok 1 - passes and logs
+  ---
+  duration_ms: 1.184595
+  type: 'test'
+  ...
+# Subtest: throws after it returned
+ok 2 - throws after it returned
+  ---
+  duration_ms: 0.748862
+  type: 'test'
+  ...
+# Subtest: waits after the throw
+ok 3 - waits after the throw
+  ---
+  duration_ms: 51.019875
+  type: 'test'
+  ...
+# Error: Test "throws after it returned" at test/b.test.ts:2:1 generated asynchronous activity after the test ended. This activity created the error "Error: the bus closed under the test" and would have caused the test to fail, but instead triggered an uncaughtException event.
+# Subtest: test/b.test.ts
+not ok 2 - test/b.test.ts
+  ---
+  duration_ms: 172.91715
+  type: 'test'
+  location: '/tmp/1901k/node/logcrash/test/b.test.ts:1:1'
+  failureType: 'testCodeFailure'
+  exitCode: 1
+  signal: ~
+  error: 'test failed'
+  code: 'ERR_TEST_FAILURE'
+  ...
+1..4
+# tests 4
+# suites 0
+# pass 3
+# fail 1
+# cancelled 0
+# skipped 0
+# todo 0
+# duration_ms 185.744334