Image sequence import improvements

[ferdnyc]

Several fixes to the process of importing image sequences:

• Also log (at warning level) the exception that caused an import failure QMessageBox
• Add paths, filenames to logging of image sequence discovery
• Reset ignored_image_sequence_paths after a successful import

[ferdnyc]

The last one bears some explaining:

The current logic around ignored_image_sequence_paths is broken and weird. It's tied to the current view object for the Project Files, so literally just switching between Thumbnail and Details view will clear out the ignored paths. The ignored paths are also cleared whenever a drag-and-drop import is done. So, they're not really stored very persistently.

But, within the current view, the list can be annoyingly persistent. For instance:

• If I import an image sequence, then remove it from Project Files and try to reimport it, that sequence won't be recognized the second time because the directory is STILL on the ignored paths list.
• Importing two different image sequences from the same directory also fails, even if they have completely distinct names. But if I switch the view from Thumbnail to Details or vice versa, then it imports fine. There's no sense to that.

So, as a result, now the list is just cleared after every successful import. Which means the user can theoretically import the same image sequence twice, if they choose to... and so what? They could always do that anyway, by switching the view or using drag-and-drop between the two import attempts.

TBH I'm tempted to just remove the ignore list entirely, as I'm pretty sure this change means that it has no actual purpose anymore. I just haven't looked at whether it might provide some protections during the sequence discovery process, that would make it still useful.

[jonoomph]

👍 Yeah, image sequences are a bit wonky. I'm tempted to re-implement them from the ground up, and use our QtImageReader to fully support them, including looping support (bounce or reset), etc... Also, support for missing files is broken when checking for image sequences.

[ferdnyc]

@jonoomph See the discussion at #221, on that front. (Wherein I try to trick a user into doing that rewrite, after they discovered that it's possible to set a start_number for FFmpeg so that the sequence isn't forced to be numbered from 0. Which is actually really great, and would improve the image sequence support greatly, but would require a complete reimplementation of the discovery/import process along the lines of what I sketch out in the last comment.)