Add windows, macOS and freebsd jobs

[ethercrow]

No description provided.

[ethercrow]

Curiously, the freebsd job complained about this:

          t_toTitle_1stNotLower:   FAIL
            *** Failed! Falsified (after 55 tests and 5 shrinks):
            "\4322"
            Use --quickcheck-replay=174726 to reproduce.

[Bodigrim]

Yeah, I've seen t_toTitle_1stNotLower test failure couple of times before :(

[ethercrow]

There's seems to be a windows specific issue:

hGetContents_crash:            FAIL
      Exception: .\cra4F80.txt: DeleteFile "\\\\?\\D:\\a\\text\\text\\cra4F80.txt": permission denied (The process cannot access the file because it is being used by another process.)

@Bodigrim didn't bytestring have something similar?

[Bodigrim]

@ethercrow sounds similar to haskell/bytestring#320

[Bodigrim]

Looks good to me, thanks!

[ethercrow]

Now we know that the benchmark suite doesn't build on Windows:

benchmarks\cbits-bench\time_iconv.c:1:10: error:
     fatal error: iconv.h: No such file or directory
        1 | #include <iconv.h>
          |          ^~~~~~~~~
  |
1 | #include <iconv.h>

  |          ^
compilation terminated.
`gcc.exe' failed in phase `C Compiler'. (Exit code: 1)

[Bodigrim]

Ahh, interesting. I think it's acceptable to remove iconv benchmark with all relevant cbits: it is not particularly important nowadays. We should not expect iconv being present out of the box on non-Windows platforms either, and I do not feel benchmarking it is worth environmental complications.

[phadej]

FWIW, iconv is always there on non-Windows platforms. GHC relies on it. https://hackage.haskell.org/package/base-4.15.0.0/docs/src/GHC-IO-Encoding-Iconv.html#iconvEncoding is one example.

In other words, you can reasonably expect iconv to be present on non-Windows platforms.

[ethercrow]

For now I just removed building benchmarks on windows.