Add new relpath filesystem

[purpleidea]

This new filesystem implements a relative filesystem modifier which can
be useful for converting between absolute filesystems rooted at / and
relative ones. This is particularly useful when interfacing with the
golang embed package.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[purpleidea]

Sorry CLA's are a rude form of asymmetrical power against contributors who are giving you something for free. There's no good reason for a CLA here.

[CrimsonFez]

Sorry CLA's are a rude form of asymmetrical power against contributors who are giving you something for free. There's no good reason for a CLA here.

Did you actually read it??? They just want you to verify that you have the rights to the code you're submitting.

[spf13]

Two issues before this can move forward:

1. CLA not signed. Please sign the CLA at https://cla-assistant.io/spf13/afero — required for all contributions.

2. TrimPrefix without HasPrefix guard. RealPath calls strings.TrimPrefix(name, r.prefix) unconditionally. If a caller passes a path that doesn't start with the configured prefix, the path is returned unchanged rather than an error being returned. There's a commented-out guard in your code suggesting you were aware of this — it should be enforced. Something like:

   if !strings.HasPrefix(name, r.prefix) {
       return name, os.ErrNotExist
   }

3. No tests. RelPathFs needs at least a table-driven test covering: paths that have the prefix (should strip), paths that don't have the prefix (should error), and Name() on a file (should add prefix back).

The concept is valid and distinct from BasePathFs (which prepends; this strips). Fix the above and it's a reasonable addition.

[spf13]

Closing with thanks — the use case you identified (bridging between absolute and relative path spaces, particularly for embed.FS) is real and valid. However, this is now better addressed by the NewBridgeIOFS adapter (#575), which lets you import any io/fs.FS directly as an afero.Fs. That's the direction the project is taking for this problem. Thanks for identifying the gap early.