Cache duplicate HTTP requests

[sirbrillig]

This is a revised version of #5 which did not actually work.

Fixes #3

To do:

• clean up the commits to remove the proxy-fetch stuff
• maybe move cache testing to the tests for getNotifications rather than on fetchNotifications?
• gettings null for the URL in gitnews-cli. Not sure why.
• maybe also include a proxy fetch so we can use the same response for duplicate HTTP requests made at the same time.

[sirbrillig]

I think I found the problem with my proxy-fetch technique. Using Response.clone() causes Response.json() to hang forever. This seems to be a known issue in node-fetch: node-fetch/node-fetch#139

I will probably need to redesign the caching system to cache decoded response data rather than the Response object itself.

[sirbrillig]

Refactored to cache responses instead.

[sirbrillig]

Rebased onto v3

[sirbrillig]

The current problem, I believe, is that groups of secondary API requests are made async, and all called at once, such that it's difficult to have the data returned by one used by the others. The requests should probably be de-duplicated by URL before they are sent out.