[attilayener]

Fixes #1743

Problem

On platforms whose file-name coding system doesn't cover some characters
(e.g. Emacs on Windows with a legacy system code page), project file names
containing non-ASCII characters are collected as raw byte escapes like
\305\264\303\251. The escapes end up both in commands such as
projectile-find-file and in the persistent cache.

Solution

As discussed in #1743, this adds a new option -
projectile-files-transform-functions - that applies a list of functions
to every file name Projectile collects for a project. The transforms run
after indexing and before the list is cached and presented, so the
cache holds the final representation as well.

The default is nil: file names are used as collected and no extra work
is done, so nothing changes for existing users.

;; example: decode raw-byte file names as UTF-8
(setq projectile-files-transform-functions
      (list (lambda (file) (decode-coding-string file 'utf-8))))

All indexing methods (alien, hybrid, native) pass through the same
transform step.

Notes

• documented in doc/modules/ROOT/pages/indexing.adoc
• covered by tests in test/projectile-indexing-test.el
• changelog entry added