repo.stat reads entire of blockstore into memory

[alanshaw]

In IPFS repo.stat takes about 5s per call for me. I found that that the entire blockstore is read into memory every time this function is called to get the count and size:

js-ipfs-repo/src/index.js

Lines 301 to 317 in 3611043

	blocks: (cb) => this.blocks.query({}, (err, list) => {
	list = list || []
	const count = new Big(list.length)
	let size = new Big(0)
	list.forEach(block => {
	size = size
	.plus(block.value.byteLength)
	.plus(block.key._buf.byteLength)
	})
	cb(err, {
	count: count,
	size: size
	})
	}),

This needs to be streamed and/or calculated once, cached, and kept up to date.

[jacobheun]

Getting that stats definitely needs to return a snapshot of the blockstore (even if it still processing). Running through the whole blockstore on each stats request is bad times.

We should probably also add some indicator that the stats are still processing while the blockstore is being read. Ideally we'd also persist this so that nodes with large blockstores that are restarted don't have to read through the entire thing if there is already a stats object in store.

[achingbrain]

This was fixed by #199 Phew!