Potential crash consistency bug in unlink

[paulwedeck]

Hello,
I discovered a similar bug to #20 but in the unlink operation.
If unlink fails after the inode has been written to the truncate list, the truncate list recovery code will "truncate" the file to its former size.
During this operation, it updates the file timestamp.

I think this is invalid because this is neither the original nor the intended final state.
My fix works by only truncating the file if the new size is different from the old size.