Remove `spec/win32_std_spec.cr`

[HertzDevil]

With just 8 out of 259 spec files failing to compile on Windows, there is no longer a dire need to keep a list of the spec files that do build. This PR makes the remaining specs use skip_file instead, similar to the Unix domain socket ones. It looks like this enables 3 new examples somewhere.

The script that generates spec/win32_std_spec.cr remains intact in case we want to reuse it for some other completely new system.

[straight-shoota]

👀

1) HTTP::Client tests write_timeout
     Failure/Error: expect_raises(IO::TimeoutError, {% if flag?(:win32) %} "WSASend timed out" {% else %} "Write timed out" {% end %}) do
       Expected IO::TimeoutError, got #<Socket::ConnectError: Error connecting to 'localhost:64862': Overlapped I/O event is not in a signaled state.> with backtrace:
         # src\socket\addrinfo.cr:67 in 'initialize'
         # src\socket\tcp_socket.cr:27 in 'initialize'
         # src\socket\tcp_socket.cr:27 in 'new'
         # src\http\client.cr:789 in 'io'
         # src\http\client.cr:671 in 'send_request'
         # src\http\client.cr:603 in 'exec_internal_single'
         # src\http\client.cr:596 in 'exec_internal'
         # src\http\client.cr:581 in 'exec'
         # src\http\client.cr:713 in 'exec'
         # src\http\client.cr:406 in 'post:body'
         # spec\std\http\client\client_spec.cr:296 in '->'
         # src\spec\example.cr:45 in 'internal_run'
         # src\spec\example.cr:33 in 'run'
         # src\spec\context.cr:18 in 'internal_run'
         # src\spec\context.cr:339 in 'run'
         # src\spec\context.cr:18 in 'internal_run'
         # src\spec\context.cr:156 in 'run'
         # src\spec\dsl.cr:220 in '->'
         # src\crystal\at_exit_handlers.cr:14 in 'run'
         # src\crystal\main.cr:50 in 'exit'
         # src\crystal\main.cr:45 in 'main'
         # src\crystal\main.cr:127 in 'main'
         # src\crystal\system\win32\wmain.cr:35 in 'wmain'
         # D:\a\_work\1\s\src\vctools\crt\vcstartup\src\startup\exe_common.inl:288 in '__scrt_common_main_seh'
         # C:\Windows\System32\KERNEL32.DLL +85712 in 'BaseThreadInitThunk'
         # C:\Windows\SYSTEM32\ntdll.dll +517019 in 'RtlUserThreadStart'

This failing spec is already activated before this PR, so it's probably not a new addition.

[HertzDevil]

I have been getting that error intermittently on local builds as well, plus this one from the first CI run:

Unhandled exception in spawn: WSASend: An established connection was aborted by the software in your host machine. (IO::Error)
  from src\io\overlapped.cr:181 in 'unbuffered_write'
  from src\io\buffered.cr:136 in 'write'
  from src\io.cr:877 in 'write_byte'
  from src\io\buffered.cr:168 in 'write_byte'
  from src\char.cr:897 in 'to_s'
  from src\io.cr:174 in '<<'
  from src\http\client\response.cr:72 in 'to_io'
  from spec\std\http\client\client_spec.cr:17 in '->'
  from src\fiber.cr:146 in 'run'
  from src\fiber.cr:98 in '->'

This blocks the whole spec. I have no idea if #12281 caused these errors or the root cause is simply an incomplete event loop implementation. I got one CI run green.

[straight-shoota]

This looks very much like a flaw in the event loop.

[HertzDevil]

The three new examples introduced in this PR are spec/std/crystal/compiler_rt/{ashlti3,ashrti3,lshrti3}_spec.cr. It is rather unlikely they would break the concurrency-related specs.

I am wondering if the failures can also be reproduced by simply rearranging the requires in win32_std_spec.cr.

[straight-shoota]

CI is green consistently, so I think this should be good to merge. If the failure shows up again, it's probably not related to this change.