proposed fix for cryfs filesystem issue

[Michael-Gallo]

For issue #1543

Works on my machine for getting rid of the high cpu usage.

I can not seem to find a way to detect a cryfs file system that doesn't rely on syscall

[joelim-work]

Even though the root cause comes from cryfs and this is merely a workaround, I think it might be OK to merge this kind of patch, since the changes are relatively small. I am not sure what other scenarios result in a block count of 0 - if it does cause other issues we will have to revert this change, but hopefully that shouldn't be the case.

One other thing is that even with this workaround, because mtime is now ignored I believe there's no way for lf to automatically reload the directory if it changes (e.g. a new file is created), so the user will have to manually call reload.

[joelim-work]

Thanks for the patch, I will give my approval.

[joelim-work]

I have bad news, it looks like tmpfs filesystems (e.g. /tmp on my machine) also have a block count of 0. Disabling nav.checkDir in this case will mean that it's impossible for lf to update the directory contents from load, period timer, etc., and the only way I can think of is to perform a full reload manually, which is not ideal.

Unless there's some other way to distinguish cryfs mounts, I don't think this kind of patch can be merged, since it will fix this bug but end up causing other issues.

Found issue with tmpfs filesystems