bug#22894: 25.1.50; dired-mark: Not remark a marked file

[Tino Calancha <f92capac@gmail.com>]

> You're right.  :-)  I think if the user tries to change the marks in a
> region, then the marks in the region should be changed.  So I'm not sure
> I understand the use case at all...
User can change the marks with:
1) `dired-change-marks': this is, IMO, the recommended way: you control 
what you want to change, let's say, the mark 'F' into the mark 'G'.
2) Calling `dired-mark-files-regexp': the patch does't prevent this 
function changing mark 'F' into `dired-marker-char'.
3) Calling `dired-flag-file-deletion', that is, changing from mark 'F'
to 'D' (that is what i called before 1) in previous comunication).
4) In two steps: unmarking the file, and after that marking it.

What i am trying to prevent is related with `dired-mark' and 
`dired-mark-files-in-region'. The former is bound to 'm'. I can imagine
someone, keeping push the 'm' button to mark a bunch of files, and 
releasing such button one fraction of seocnd late, so one marked file with
'F' get remarked by `dired-marker-char'.
Similar thing could happen if using the second function setting the region 
not very carefully (pick uping one additional file up/down in the region).

The patch just prevent in this two function, one marked file be remarked. 
Those files still not marked are marked.

I am against to restrict users, and i understand this thread is 
controversial, but i use a lot these features and i believe it could
prevent people (including me) doing unintentional changes.