Fix log.tail by calling add after listening for events

[magik6k]

This got broken in go-ipfs 0.4.18, I have no idea why it worked before..

[achingbrain]

There are some linting failures:

1. Strings should use single quotes
2. The commit message should not be sentence case so I think fix: Fix log.tail.. needs changing to be fix: fix log.tail

[daviddias]

@magik6k rebased master onto your branch so that it runs with latest go-ipfs.

Note, I didn't see it fail on master.

[alanshaw]

I'm not sure if I'm mis-interpreting what is happening here. My understanding of this is:

1. We setup an interval to run once every second
2. Call ipfs.log.tail - which will callback almost immediately with a stream for us to consume
3. Clear the interval - presumably before the first time it runs (unless ipfs.log.tail takes > 1s to callback)

So I think adding this interval is either ineffectual, or it's relying on ipfs.log.tail to take > 1s to callback.

If this test is passing then it might be a false positive, which we need to get to the bottom of.

[magik6k]

Yeah, this should be in res.once callback.. It was quite late when I wrote this fix.

Looked at this more - with ipfs daemon --offline, doing curl http://127.0.0.1:5001/api/v0/log/tail -v, go-ipfs will not send response headers until first log message, so this is why this worked.

This test also works currently because go-ipfs tends to report some random events, but there is no guarantee that they will happen and this causes the test to hang for way too long.

[alanshaw]

Merging as test failures are for ping tests - unrelated to this PR.