Attempted fix for urllib3 crashes

[sg3-141-592]

We've still got an issue with crashes on the urllib3 requests test that uses the mock HTTP server.

Fix #9958 to handle port mapping errors didn't resolve it.

I got a feeling there's an ordering issue. Looking at the error logs https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=56500#c2 there appears to be an issue where we're throwing exceptions before the coverage completes.

=== Uncaught Python exception: ===
--
  | MaxRetryError: HTTPConnectionPool(host='localhost', port=8011): Max retries exceeded with url: / (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7f4cdf33d1f0>: Failed to establish a new connection: [Errno 101] Network is unreachable'))
  | Traceback (most recent call last):
  | File "fuzz_requests.py", line 109, in TestOneInput
  | File "urllib3/_request_methods.py", line 118, in request
  | File "urllib3/_request_methods.py", line 217, in request_encode_body
  | File "urllib3/poolmanager.py", line 433, in urlopen
  | File "urllib3/connectionpool.py", line 874, in urlopen
  | File "urllib3/connectionpool.py", line 874, in urlopen
  | File "urllib3/connectionpool.py", line 874, in urlopen
  | File "urllib3/connectionpool.py", line 844, in urlopen
  | File "urllib3/util/retry.py", line 505, in increment
  | MaxRetryError: HTTPConnectionPool(host='localhost', port=8011): Max retries exceeded with url: / (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7f4cdf33d1f0>: Failed to establish a new connection: [Errno 101] Network is unreachable'))
  |  
  | INFO: Instrumenting 3854 functions...
  | INFO: Instrumentation complete.
  | ==10674== ERROR: libFuzzer: fuzz target exited
  | #0 0x7f4ce0bac694 in __sanitizer_print_stack_trace /src/llvm-project/compiler-rt/lib/ubsan/ubsan_diag_standalone.cpp:31:3
  | #1 0x7f4ce0b2df48 in fuzzer::PrintStackTrace() /src/llvm-project/compiler-rt/lib/fuzzer/FuzzerUtil.cpp:210:5
  | #2 0x7f4ce0b12cdc in fuzzer::Fuzzer::ExitCallback() /src/llvm-project/compiler-rt/lib/fuzzer/FuzzerLoop.cpp:250:3
  | #3 0x7f4ce09068a6 in __run_exit_handlers /build/glibc-SzIz7B/glibc-2.31/stdlib/exit.c:108:8
  | #4 0x7f4ce0906a5f in exit /build/glibc-SzIz7B/glibc-2.31/stdlib/exit.c:139:3
  | #5 0x7f4ce03b2c78 in libpython3.8.so.1.0
  | #6 0x7f4ce03b76cf in libpython3.8.so.1.0
  | #7 0x403ad2 in fuzz_requests.pkg
  | #8 0x403e67 in fuzz_requests.pkg
  | #9 0x7f4ce08e4082 in __libc_start_main /build/glibc-SzIz7B/glibc-2.31/csu/libc-start.c:308:16
  | #10 0x40249d in fuzz_requests.pkg
  |  
  | SUMMARY: libFuzzer: fuzz target exited

This is an attempted fix inspired by the requests fuzz_server.py where the lifecycle of the test thread is managed within the server. Since the web server is created at the start of TestOneInput I don't expect there to be any timing issues or thread initialisation issues.