move the bookmark validation logic into the model, as well as the
bookmark to root code

if #delete() was called on a bookmark that wasnt cached, the wrong error
was being "catch"ed

now we dont assume the cursor is in the nerd tree window when calling
s:SaveScreenState()

calling filereadable() on a FIFO file never terminates so we have to
check for them

invalid bookmarks now placed after a blank line at the bottom of the
bookmarks file

if invalid bookmarks are detected, an error is outputted which directs
the user to :help NERDTreeInvalidBookmarks, which had been added to the
help file in this commit

add a '^' and '$' to the end of the file matching pattern so it cant get
confused by buffer names that happen to be a substring of buffer we are
comparing against

previously the mouse handling function was not passing the needed
argument to the ActivateNode function

now if invalid bookmarks are detected, the offending lines are moved to
the bottom of the bookmarks file.

if the user tries to activate a bookmark that has become invalid after
it was read, that bookmark will be removed from the cache and put at the
bottom of the bookmarks file

This reverts commit 21601d9c3d.

Ok, so that copying command doesnt work for windows... it doesnt do
directories properly.

we dont need it to default to 1 anymore as that was a hack for windows,
but now that we handle windows paths properly we dont need it

when the user hits B and opens the bookmarks table, stick the cursor at
the top of it

this way feels cleaner

make the t/T on directory nodes open a fresh NERD tree for the selected
dir in a new tab, rather than a netrw.

doesnt work for paths with some screwed up chars in them

when opening a file in a new split (with the <tab> map) we were doing
the :split with a "silent" modifier. This meant that a swap existed and
input was required from the user, the script would just freeze up