Basic torrent persistence

[njoyard]

Here is my take on torrent persistence.

From a user's point of view here's how it is intended to be used:

var data = client.getPersistData();
// Store data somewhere, stop the client
// (any further progress won't be saved of course)

// ... later retrieve the data, create a new client and restore torrents
var torrents = client.restorePersistData(data);

getPersistData returns an object holding persistent data (including some Buffers). It includes the following for every torrent:

• download path
• bitfield
• tracker announce urls
• piece bitfields
• metadata (infoHash, bitfield and encoded metadata)

I guess it is up to the user to ensure this data is saved and restored consistently. I'm not sure whether Buffers should be serialized by node-torrent or by the user. I'm also not sure whether it would be a good idea to persist peer information to speed up torrent restarts.

Upon restoring, invalid pieces are cleared.

I was not able to test this a lot, so this pull request is mainly to get opinions on the implementation for now.