* wip: skeleton code for trash adapter
* refactor: split trash implementation for mac and linux
* fix: ensure we create the .Trash/$uid dir
* feat: code complete linux trash implementation
* doc: write up trash features
* feat: code complete mac trash implementation
* cleanup: remove previous, terrible, undocumented trash feature
* fix: always disabled trash
* feat: show original path of trashed files
* doc: add a note about calling actions directly
* fix: bugs in trash implementation
* fix: schedule_wrap in mac trash
* doc: fix typo and line wrapping
* fix: parsing of arguments to :Oil command
* doc: small documentation tweaks
* doc: fix awkward wording in the toggle_trash action
* fix: warning on Windows when delete_to_trash = true
* feat: :Oil --trash can open specific trash directories
* fix: show all trash files in device root
* fix: trash mtime should be sortable
* fix: shorten_path handles optional trailing slash
* refactor: overhaul the UI
* fix: keep trash original path vtext from stacking
* refactor: replace disable_changes with an error filter
* fix: shorten path names in home directory relative to root
* doc: small README format changes
* cleanup: remove unnecessary preserve_undo logic
* test: add a functional test for the freedesktop trash adapter
* test: more functional tests for trash
* fix: schedule a callback to avoid main loop error
* refactor: clean up mutator logic
* doc: some comments and type annotations
If you rename a file to have a `/` at the end, oil will detect the
change as a delete + new directory. Similarly, if you delete the
trailing `/` from a directory, it will delete the dir and create a new
file. This should help with the case where you want to create a new file
but forgot to add the trailing `/` and now you have a file instead.
Previously we were gc-ing all hidden oil buffers 2 seconds after they
were hidden. I think this is a little too magical, and interferes with
some expected vim behavior (ctrl-o/i). If people want the old behavior,
we can expose the "GC hidden buffers" function via the API.
We're also changing the "rerender visible oil buffers" logic, because
previously that would delete hidden oil buffers. Now it simply marks
them as dirty, and they will be rerendered during the next BufEnter.
