CLI: Remove use of output "prefixes"

[Lestropie]

Related to #3152.

For interfacing with other software ecosystems as opposed to human users, it would be preferable that if a command generates more than one file based on a user-specified path, particularly if that set of files is not necessarily known at compile time, that the command interface require that the user specify an output directory, rather than a string prefix. This would expedite parsing of the command outputs, and also simplify checking of pre-existing filesystem data that clashes with prospective outputs.

Going to attempt here to generate a list of command interfaces that I think should be modified as a single cohesive change.

Fixel command interface

It's possible that this target might not be possible without either addressing #1043 or doing some other workaround.
Many fixel-related commands take the input and output fixel directories as separate command-line options.
If the output directory does not exist, then it is created, and index.mif and directions.mif are copied from the input fixel directory.
However .type_directory_out() enforces that the output directory must not exist.
Even with -force, it will refuse to delete an entire pre-existing directory.
Therefore these commands all have the output directory specified as .type_text(), to avoid the overwrite check.

I see two solutions:

1. Change .type_directory_out() to accept a bollean specifying whether the directory already existing is OK.
2. Change all fixel-related commands to have the output fixel data file path as a path in its own right, saving it to whether the user specifies. If the user wants that file to go into a new fixel directory, that has the same index.mif and directions.mif as another fixel directory, then they can make that copy themselves. Personally however I think this is bad practise: there should be no harm in a fixel directory having a new fixel data file created within it, and having multiple fixel directories with exactly the same fixes is redundant. So this would be my preference.

Changelist

• fixelcfestats: Output fixel directory should be .type_directory_out() rather than .type_text()` (no actual change to behaviour).

• tck2fixel: Output directory and fixel data file name both specified as .type_text() (see "Fixel command interface" above)

• 5ttcheck: Change -voxels option so that if multiple input images are provided, it creates a directory in which to store all of the respective masks.

• mrclusterstats: Change output from a prefix to a directory.

• connectomestats: Change output from a prefix to a directory.

• voxel2fixel: See "Fixel command interface" above.

• connectomeedit: Change input & output to files rather than text.

• warp2metric: See "Fixel command interface" above.

• fixelcorrespondence: See "Fixel command interface" above.

• vectorstats: Change output from a prefix to a directory.

• connectome2tck: If output is not a single tractogram file, create a directory with the path specified by the user.

• fixelconvert: Remove command I think.

• tsfinfo:

  • -ascii option should be an output directory rather than a prefix.
  • Consider moving this out of tsfinfo. tsfconvert would make more sense.

• tckedit / tckgen: -include_ordered is currently .type_text() whereas others are .type_various(). Ideally all of these would change as per API: Replace .type_various() #3074.

• mrregister: -linstage.diagnostics.prefix change to output directory.

• mrview: -capture.folder change from .type_text() to .type_directory_out().

• fixelcfestats: Change input fixel-fixel connectivity matrix to be .type_directory_in() (should no longer need to catch pre-3.0.0 usage attempts)

• Look more closely for anything in the Python commands;
  may be something in there that's an implicit string that I won't find with grep?