Simplify cleanup on failure in realpath(3).

[kib]

Summary

If realpath() allocated memory for result and failed, the memory is freed in each place where return is performed. More, the function needs to track the allocation status, to not free user-supplied buffer.

Consolidate the memory handling in the wrapper, freeing the buffer if the actual worker failed.

This conflicts with the patch in PR 219154, I will pass the PR change first.

[kib]

Do malloc() in realpath(), as before. Pass only 'resolved' to realpath1(), all logic to malloc and free is now located in realpath().

The conditional recording of "." path is now done always, in worst case of malloced result this two-bytes write is lost.

In principle, now realpath1() arguments could be qualified with restrict, but I decided no to, since 'sufficiently smart compiler' should be able to figure on its own, and restrict would make a difference only if compiler is smart enough.

Suggested by: emaste

[kib]

Update after merge with r318298.