Better error handling for when request function throws

[lukechilds]

Related: sindresorhus/got#392

When the request function throws (as opposed to emitting an error) it causes cacheable-request to emit an error. This is misleading because errors emitted by cacheable-request are assumed to be caused by a cache error, request errors are handled by listening for the error event on the request object.

However, we can't attach an event listener to the request object in this scenario because the request object wont exist if the request function throws.

We need a clearer way to handle this error.

[lukechilds]

Possible solution: cacheable-request should have two error classes for RequestError and CacheError.

So you could handle all possible errors from cacheable-request like this:

cacheableRequest('example.com', cb)
  .on('error', err => {
    if (err instanceof cacheableRequest.CacheError) {
      handleCacheError(err); // Cache error
    } else if (err instanceof cacheableRequest.RequestError) {
      handleRequestError(err); // Request fn thrown
    }
  })
  .on('request', req => {
    req.on('error', handleRequestError); // Request error emitted
    req.end();
  });