fix: improve Fish shell support (and Cshell)

[mtb0x1]

I faced the same issue mentioned in #1299, this PR fixes the issue by taking in account specific set and unset syntax for Fish shell.

The post install instruction (i.e : ./emsdk activate latest) provides wrong instructions in Fish shell and Cshell cases.
This should be fixed by this PR too, although the code is a bit redundant and it could be more cleaner.

The website instruction have the same issue, but I guess user with specific shell will be able to manage.

[mtb0x1]

I am not sure why the test-linux ci failed, doesn't seem to be related to this PR.
Running the failed step ( test_binaryen_from_source) manually on my end works fine. a Ci timeout ?

[sbc100]

Thanks for working on this. One last nit and we can submit I think

[sbc100]

Looks like flake8 also found a couple of nits

[mtb0x1]

Looks like flake8 also found a couple of nits

Yes, sorry about that I don't have a proper python setup locally I have to check ci output everytime.
I will let you know when it's passing.

[mtb0x1]

flake8 is fine now, but test-linux is failing again.
I am not sure of it's a timeout or a real issue that I am missing.

logs in https://circleci.com/api/v1.1/project/github/emscripten-core/emsdk/14456/output/106/0?file=true&allocation-id=66aabe4cea13c3196692b681-0-build%2FABCDEFGH, indicate

...
[ 14%] Building CXX object src/wasm/CMakeFiles/wasm.dir/parsing.cpp.o
c++: fatal error: Killed signal terminated program cc1plus
compilation terminated.
make[2]: *** [src/parser/CMakeFiles/parser.dir/build.make:128: src/parser/CMakeFiles/parser.dir/parse-3-implicit-types.cpp.o] Error 1
make[2]: *** Waiting for unfinished jobs
...

I am guessing it's a resource issue, the step test_binaryen_from_source is building binaryen from source with all available cores (in this case 35) and somehow this causes a OOM ?

[sbc100]

Oops, looks like there was a missing comma: #1433