generic: test bdev freeze count leak on nested thaw

Since kernel v6.8, a nested thaw that drops its freeze reference
while other freezers remain returns -EINVAL although it succeeded.
bdev_thaw() then keeps bd_fsfreeze_count elevated, and since
device-mapper's unlock_fs() ignores the error, a dm suspend/resume
cycle on a filesystem frozen with FIFREEZE leaks the count. After
unfreezing and unmounting, the block device can never be mounted
again:

    dm-1: Can't mount, blockdev is frozen

Add a regression test that freezes a filesystem on a dm-linear
device, runs a dm suspend/resume cycle inside the freeze, unfreezes
and remounts. On buggy kernels the final mount fails with EBUSY.

Suggested-by: Christoph Hellwig <hch@lst.de>
Lore: https://lore.kernel.org/linux-fsdevel/20260821085451.65206-1-moritz.tanner@linbit.com/
Signed-off-by: Moritz Tanner <moritz.tanner@linbit.com>
Reviewed-by: Zorro Lang <zlang@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Zorro Lang <zlang@kernel.org>