Fix Windows only problem of wrapper command built by npm

[taiju]

rel: #34, #69, #138

-- I'm sorry. I'm not good at English.

psc-package command can't exec purs, if it is installed by npm command.

Although I tried to solve it, I found that it is a constraint of System.Process.createProcess or System.Process.runProcess.

• createProcess function is can't exec a *.cmd or *.bat file haskell/process#140

It works if instead use System.Process.shell or System.Process.runCommand.

This PR is another approach of #69.

Thank you.

[justinwoo]

This makes psc-package build not work for me on Linux, where I have a purs binary only. Problem seems to be from the source globs not being passed in correctly.

[justinwoo]

E.g. one test run I tried results in this:

psc-package: ShellFailed {shellCommandLine = "purs compile .psc-package/testing/arrays/v5.1.0/src/**/*.purs .psc-package/testing/bifunctors/v4.0.0/src/**/*.purs .psc-package/testing/control/v4.1.0/src/**/*.purs .psc-package/testing/distributive/v4.0.0/src/**/*.purs .psc-package/testing/effect/v2.0.0/src/**/*.purs .psc-package/testing/either/v4.1.0/src/**/*.purs .psc-package/testing/enums/v4.0.0/src/**/*.purs .psc-package/testing/exceptions/v4.0.0/src/**/*.purs .psc-package/testing/foldable-traversable/v4.1.0/src/**/*.purs .psc-package/testing/foreign/v5.0.0/src/**/*.purs .psc-package/testing/foreign-object/v1.1.0/src/**/*.purs .psc-package/testing/functions/v4.0.0/src/**/*.purs .psc-package/testing/gen/v2.1.0/src/**/*.purs .psc-package/testing/globals/v4.0.0/src/**/*.purs .psc-package/testing/identity/v4.1.0/src/**/*.purs .psc-package/testing/integers/v4.0.0/src/**/*.purs .psc-package/testing/invariant/v4.1.0/src/**/*.purs .psc-package/testing/lazy/v4.0.0/src/**/*.purs .psc-package/testing/lists/v5.3.0/src/**/*.purs .psc-package/testing/math/v2.1.1/src/**/*.purs .psc-package/testing/maybe/v4.0.0/src/**/*.purs .psc-package/testing/newtype/v3.0.0/src/**/*.purs .psc-package/testing/nonempty/v5.0.0/src/**/*.purs .psc-package/testing/nullable/v4.1.0/src/**/*.purs .psc-package/testing/orders/v4.0.0/src/**/*.purs .psc-package/testing/partial/v2.0.0/src/**/*.purs .psc-package/testing/prelude/v4.1.0/src/**/*.purs .psc-package/testing/proxy/v3.0.0/src/**/*.purs .psc-package/testing/record/v1.0.0/src/**/*.purs .psc-package/testing/refs/v4.1.0/src/**/*.purs .psc-package/testing/simple-json/v4.3.0/src/**/*.purs .psc-package/testing/st/v4.0.0/src/**/*.purs .psc-package/testing/strings/v4.0.1/src/**/*.purs .psc-package/testing/tailrec/v4.0.0/src/**/*.purs .psc-package/testing/transformers/v4.1.0/src/**/*.purs .psc-package/testing/tuples/v5.1.0/src/**/*.purs .psc-package/testing/type-equality/v3.0.0/src/**/*.purs .psc-package/testing/typelevel-prelude/v3.0.0/src/**/*.purs .psc-package/testing/unfoldable/v4.0.0/src/**/*.purs .psc-package/testing/unsafe-coerce/v4.0.0/src/**/*.purs .psc-package/testing/variant/v5.2.0/src/**/*.purs", shellExitCode = ExitFailure 1}

Maybe the problem arises from these source globs not being quoted?

[justinwoo]

It seems you'd have to fiddle with both String and Text related code to quote all of these (e.g. justinwoo@e008bd9), and it still feels a bit iffy as you can pass through options to the compiler. Maybe someone will have another opinion about this.

[taiju]

Well, I see...
I will try another approach.
Do you think that it is worth keeping this approach?

[taiju]

I understood that there are many problems this PR. I'll rethink. Thank you for your reviewing!