Implement --md5 and --channel-subdir for non-explicit env export

[jonashaag]

This keeps the old behaviour of defaulting to md5 output for explicit files and no md5 output for environment files, but allows for outputting md5 with --md5 for environment files.

[JohanMabille]

I've just merged the fix for the umamba_tests_win_cmd ci job in the main branch, rebasing your PR should make the CI pass.

[jonashaag]

Done, I've also added --channel-subdir

[jonashaag]

build/micromamba/micromamba env export -p /tmp/env --channel-subdir --md5
name:
channels:
- conda-forge
dependencies:
- conda-forge/osx-arm64::bzip2=1.0.8=h3422bc3_4[md5=fc76ace7b94fb1f694988ab1b14dd248]
- conda-forge/osx-arm64::ca-certificates=2023.5.7=hf0a4a13_0[md5=a8387be82224743cf849fb907790b91a]
- conda-forge/osx-arm64::libexpat=2.5.0=hb7217d7_1[md5=5a097ad3d17e42c148c9566280481317]
- conda-forge/osx-arm64::libffi=3.4.2=h3422bc3_5[md5=086914b672be056eb70fd4285b6783b6]
- conda-forge/osx-arm64::libsqlite=3.42.0=hb31c410_0[md5=6ae1bbf3ae393a45a75685072fffbe8d]
- conda-forge/osx-arm64::libzlib=1.2.13=h53f4e23_5[md5=1a47f5236db2e06a320ffa0392f81bd8]
- conda-forge/noarch::multiregex=2.0.1=pyhd8ed1ab_0[md5=fc050ddbb61645437f080e9323d94983]
- conda-forge/osx-arm64::ncurses=6.4=h7ea286d_0[md5=318337fb9d0c53ba635efb7888242373]
- conda-forge/osx-arm64::openssl=3.1.1=h53f4e23_1[md5=7451b96ed28b5fd02f0df32689327755]
- conda-forge/noarch::pip=23.1.2=pyhd8ed1ab_0[md5=7288da0d36821349cf1126e8670292df]
- conda-forge/osx-arm64::pyahocorasick=2.0.0=py311he2be06e_0[md5=4c4e5d758113c1339abf6fe45cad6008]
- conda-forge/osx-arm64::python=3.11.4=h47c9636_0_cpython[md5=b790b3cac8db7bdf2aaced9460bdbce4]
- conda-forge/osx-arm64::python_abi=3.11=3_cp311[md5=e1586496f8acd1c9293019ab14dbde9d]
- conda-forge/osx-arm64::readline=8.2=h92ec313_1[md5=8cbb776a2f641b943d413b3e19df71f4]
- conda-forge/noarch::setuptools=68.0.0=pyhd8ed1ab_0[md5=5a7739d0f57ee64133c9d32e6507c46d]
- conda-forge/osx-arm64::tk=8.6.12=he1e0b03_0[md5=2cb3d18eac154109107f093860bd545f]
- conda-forge/noarch::tzdata=2023c=h71feb2d_0[md5=939e3e74d8be4dac89ce83b20de2492a]
- conda-forge/noarch::wheel=0.40.0=pyhd8ed1ab_0[md5=49bb0d9e60ce1db25e151780331bb5f3]
- conda-forge/osx-arm64::xz=5.2.6=h57fd34a_0[md5=39c6b54e94014701dd157f4f576ed211]

[Hind-M]

In which scenario no_md5 would be equal to -1 if it's configured as a flag and initialized to 0?
Is there any test for this use case?

[jonashaag]

I think when you pass --md5

[Hind-M]

--md5 is just a negation for no_md5 and they are both flags, so I don't think it's ever going to be equal to -1. We could either use add_option instead of add_flag to allow the user to set it to -1 or other int and have that use case, or leave it as a bool and not have that use case.

[AntoinePrv]

I'm also a bit confused why no_md5 is not a boolean. If there are more complex cases than boolean, I think an enum would be more readable.

[jonashaag]

I just experimentally verified the following behavior:

• Passing nothing: no_md5=0
• Passing --no-md5: no_md5=1
• Passing --md5: no_md5=-1

[jonashaag]

I'm merging this for now but I'm happy to improve on it using an enum or something else in a follow up PR