When paths are different in gcno/gcda archives then try to match file name + stamp

[calixteman]

Normally we expect to have a/b/c/foo.gcno and a/b/c/foo.gcda in archives but for any reason the base paths can be different.
So the idea is to use filename + stamp (or checksum for llvm gcno/gcda) as a unique identifier when paths don't match.
In case where filename + stamp doesn't identify an unique stem then we fail.

[codecov-io]

Codecov Report

Merging #400 into master will increase coverage by 0.07%.
The diff coverage is 65.71%.

@@            Coverage Diff             @@
##           master     #400      +/-   ##
==========================================
+ Coverage    65.2%   65.28%   +0.07%     
==========================================
  Files          12       12              
  Lines        4078     4162      +84     
  Branches     1010     1049      +39     
==========================================
+ Hits         2659     2717      +58     
+ Misses        649      632      -17     
- Partials      770      813      +43

Impacted Files	Coverage Δ
src/producer.rs	83.93% <65.71%> (-1.28%)	⬇️
src/lib.rs	62% <0%> (-3.5%)	⬇️
src/parser.rs	58.72% <0%> (-1.21%)	⬇️
src/reader.rs	56.42% <0%> (-0.58%)	⬇️
src/path_rewriting.rs	70.66% <0%> (-0.45%)	⬇️
src/output.rs	72.39% <0%> (+5.73%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 27980cb...5467910. Read the comment docs.

[make-github-pseudonymous-again]

Does this fix what I am experiencing here: travis lists a bunch of gcna/gcdo files covering all executables and libraries in the zip archive but codecov only lists the library source files.

[marco-c]

@calixteman do we still want to do this?

[marco-c]

If we still want to do this, the first step will be to rebase it.