Currently, I believe that when Tricium triggers swarming tasks, it doesn't request for any named cache directories to be used, although maybe we might want to use a cache for git data that is repeatedly fetched.

Currently in GitFileIsolator tasks, we do:
 1. Make a temp directory
 2. Call `git init`
 3. Call `git fetch --depth=1 --no-tags --no-recurse-submodules <repo> <ref>`
 4. Call `git checkout FETCH_HEAD -- <files...>

Example:
https://chromium-swarm.appspot.com/task?id=4122eebd6b23fa10&refresh=10&show_raw=1

AFAIU, the git fetch part could potentially be sped up if a named cache was used. Now, within our one tricium pool we checkout various different repos, so that might make this less effective in practice.