fix: do not panic on binrw backtrace error

[vthib]

The ntfs crate could panic on some data because of a false "unreachable" instruction in error.rs: the binrw error is not guaranteed to only be an IO error, it can also be for example a backtrace error.

I have reached this panic several times in different contexts.

This PR removes this unreachable which is too dangerous: any update of binrw could make this unreachable wrong, it is better to be conservative here.

[ColinFinck]

Thanks for reporting! This happened when I exchanged the abandoned binread dependency by its successor binrw. Backtrace didn't exist back in binread times.

Looking at Backtrace, it's a heavyweight error type that requires heap allocations and carries a full-blown backtrace. That wouldn't be a problem when used in an std context, but it can be a problem when used in a no_std context. I only want to propagate that Backtrace error via my NtfsError when it's absolutely necessary. And it seems like that's only the case when the "verbose-backtrace" feature of binrw is enabled.

Checking with the binrw authors on Discord if they can confirm this and what's the right way to proceed here.