dosbox-staging@0.82.0: Persistence of auto-mounted drives

[farsil]

There are three possibles approaches concerning the persistence of auto-mounted drives:

1. (current) Do not do anything. Let the user set up their filesystem structure the way they like.
2. Persist drives. This works from any folder because dosbox-staging always looks in its own folder before checking in the current directory. I've tested running the dosbox-staging executable from a different drive via the command line and it detects the drives folder without issues. I haven't seen any mention in the documentation of this behavior (but I'm asking upstream devs).
3. Persist resources/drives. I don't like this solution because the files in resources/drives may be updated later by upstream developers, but it does allow to exploit a documented feature, and allows customization of the default "ESSENTIALS" drive.

Documentation of the automount feature is present in the default configuration file:

#                   automount: Mount 'drives/[c]' directories as drives on startup, where [c] is a lower-case
#                              drive letter from 'a' to 'y' (enabled by default). The 'drives' folder can be
#                              provided relative to the current directory or via built-in resources.
#                              Mount settings can be optionally provided using a [c].conf file along-side
#                              the drive's directory, with content as follows:
#                                [drive]
#                                type    = dir, overlay, floppy, or cdrom
#                                label   = custom_label
#                                path    = path-specification, ie: path = %path%;c:\tools
#                                override_drive = mount the directory to this drive instead (default empty)
#                                verbose = true or false
#                                readonly = true or false

Personally, I would like to have 2 implemented.

[farsil]

I confirmed with upstream devs that the behavior described in 2 can be relied upon.

[farsil]

I prepared this PR (which got closed for some reasons) that implements 2

#14793