Make code more verbose and easy to read

This reverts commit 03ae4817.

Fix regression introduced in 95f14593

ConfigControl::getName() uses memory of p_item which is a part of p_config memory. This was causing a crash when trying to save the preferences, as the constructor had already destroyed it upon creation.

Instead, free the p_config memory on the destructor, which is automatically called when the dialog closes, saved or cancelled. Verified to not leak any memory.

Close #9214
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Close #9200

Without this fix, they are indistinguishable in the GUI and the user cannot know which is which.
Pointed-out-by: Eric Hameleers <alien@slackware.com>
Signed-off-by: Rafaël Carré <funman@videolan.org>

Normalise it with other options in advanced prefs.
Signed-off-by: Rafaël Carré <funman@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Close #9209

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

No longer needed as the model now behaves correctly.
reverts 442bad11

- popup now created according to the selected items and models.
- fixes view/popup model abstraction.
- allows introducing new methods/models.

Never ends..

- Separate common methods into interface, allowing multiple
inheritance architecture (to use any implemented AbstractModel)
- Code factorization between PL/ML models (and PL/ML Items)
- Use proxy model for views and redirection to the PL/ML models.
Fixes filtering and also really make use of the current model (only the
PL Model was really used before, models were half switched).
- stop storing/serializing selection indexes
- ML Missing locks and bug fixes
- Move ML callback into event loop
- ML Items leaks fixes
- Code factorization
- Dead code removal

It might be better to convert from the proper character encoding to
UTF-8, but I do not know what proper is.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

tested with ticket 9022 samples to 2 channel and without channel
definition. Output files seemed to have correct channel amount in both
cases.

refs #9022