Seeding re-stores chunks even if they are already stored

[DiegoRBaquero]

When you seed a file, it doesn't matter if the pieces are already on the store, it will create a ChunkStoreWriteStream and re-store them.

The verifying happens in Torrent._onMetadata, called on Torrent._onSwarnListening
The re-storing happens in Torrent.load, called in WebTorrent.seed

webtorrent:torrent got metadata +5ms
webtorrent:torrent verifying existing torrent data +3ms
webtorrent:torrent piece verified 0 +34ms
webtorrent:torrent piece verified 1 +1ms
webtorrent:torrent piece verified 2 +1ms
webtorrent:torrent done verifying +2ms
webtorrent:torrent on store +1ms
webtorrent:torrent select 0-2 (priority 0) +1ms
webtorrent:torrent file done: PMC-InformelaMuestra.pdf +3ms
webtorrent:torrent torrent done: 307f97a2e9264a5c17d04dd83aae6069d249a287 +1ms
bittorrent-tracker send `complete` +0ms
function onData(chunk) {
    if (self.destroyed) return
    store.put(index, chunk)
    index += 1
  }
put 0 from function (index, buf, cb) {
  console.info(arguments.callee.caller.toString())
  var self = this
  self.mem[index] = buf
  self.store.put(index, buf, function (err) {
    self.mem[index] = null
    if (cb) cb(err)
  })
}
function onData(chunk) {
    if (self.destroyed) return
    store.put(index, chunk)
    index += 1
  }
put 1 from function (index, buf, cb) {
  console.info(arguments.callee.caller.toString())
  var self = this
  self.mem[index] = buf
  self.store.put(index, buf, function (err) {
    self.mem[index] = null
    if (cb) cb(err)
  })
}
function onData(chunk) {
    if (self.destroyed) return
    store.put(index, chunk)
    index += 1
  }
put 2 from function (index, buf, cb) {
  console.info(arguments.callee.caller.toString())
  var self = this
  self.mem[index] = buf
  self.store.put(index, buf, function (err) {
    self.mem[index] = null
    if (cb) cb(err)
  })
}
webtorrent:torrent torrent done: 307f97a2e9264a5c17d04dd83aae6069d249a287 +54ms
bittorrent-tracker send `complete` +1ms

[feross]

You're right. We can make two improvements here:

1. Eliminate the verification step for ALL input types to client.seed, since it'll obviously be correct.
2. Eliminate the call to torrent.load (streaming into the store) when input is a filesystem path.