Deduplicating OnDisk Corpus

[domenukk]

Instead of having multiple files for the same content, the corpus should behave as follows:

• On add: Create a file where filename == hash(contents)
• Create a second hidden file with the same name (starting with a dot). This file should contain a simple counter. Accessing this file should happen with exclusive access (FLOCK/file lock).
• When we add a testcase where the correct file already exists, increase the counter
• When we remove this from the corpus, reduce the counter.
• If the counter reaches 0, remove the file before dropping the lock on the shadow/counter file.
• When the file contents change: instead, remove & create a new file using the algorithm above.

For reading, this shouldn't have much overhead, and it'll get rid of duplicate files with the same content! Great success :)