Work on copy of command to avoid side effect from tokenization

[tobiasfrejo]

Due to tokenization of the parsed command in slash_execute, the string gets mutated on execution.

Example:

cmd = "node 12"
pycsh.slash_execute(cmd)
print(cmd.encode())

prints b'node\x0012' instead of the expected b'node 12'

This is resolved here by copying the command string before it is passed to the slash_execute C function.

[kivkiv12345]

Good catch!
Before merging however, I'm wondering if we should introduce a verbose argument to control the new "executing" printf.
I don't now much about how this function will be used, but I imagine that the print may become tedious.
On that thought, I should mention that this function is quite divisive. I defend its inclusion, but I can't promise it will stick around.

Another thing, PyCSH has the CLEANUP_FREE macro (from "utils.h"), which I prefer to use for pointers that should be freed when they fall out of scope.
It's not too important on cmd_cpy here, since there aren't any guard clauses where we can forget to free it.

[tobiasfrejo]

You are right about the printf potentially being tedious. That was just used for debugging, so I have simply removed it now, and also added the CLEANUP_FREE for good measure.

As a defense for keeping the function, I see it as a way to call CSH commands that has not been implemented as Python functions, which we might use in DISCO for automating calls to our own libraries and not having to implement them in Python as well.

[kivkiv12345]

To my dismay, GCC is overly strict about the types of cleanup routines. So using CLEANUP_FREE instead of CLEANUP_STR to free a char* will issue a warning, even though they both call free().
That is to say, you can replace CLEANUP_FREE with CLEANUP_STR, and the warning will go away. Calling free() explicitly here is also fine by me.