chore: remove node buffers from runtime code

Node `Buffers` are subclasses of `Uint8Array` and we don't use any
`Buffer`-specific functions, so do not demand `Buffer`s where
`Uint8Array`s will do.

`Buffer`s are still used in the tests because the `Buffer` requirement
needs pull out of (at least) the `cids` and `multihash` modules first.

lerna.json

@@ -1,5 +1,5 @@
 {
-  "lerna": "3.20.2",
+  "lerna": "3.22.1",
   "packages": [
     "packages/*"
   ],

package.json

@@ -19,7 +19,7 @@
     "update-contributors": "aegir release --lint=false --test=false --bump=false --build=false --changelog=false --commit=false --tag=false --push=false --ghrelease=false --docs=false --publish=false"
   },
   "devDependencies": {
-    "lerna": "^3.20.2"
+    "lerna": "^3.22.1"
   },
   "repository": {
     "type": "git",

packages/ipfs-unixfs-exporter/README.md

@@ -50,7 +50,7 @@ const files = []
 
 for await (const file of importer([{
   path: '/foo/bar.txt',
-  content: Buffer.from(0, 1, 2, 3)
+  content: new Uint8Array([0, 1, 2, 3])
 }], ipld)) {
   files.push(file)
 }
@@ -74,7 +74,7 @@ for await (const buf of entry.content({
   bytes.push(buf)
 }
 
-const content = Buffer.concat(bytes)
+const content = new Uint8Array(bytes)
 
 console.info(content) // 0, 1, 2, 3
 ```
@@ -185,7 +185,7 @@ for await (const chunk of entry.content({
 }
 
 // `data` contains the first 5 bytes of the file
-const data = Buffer.concat(bufs)
+const data = new Uint8Array(bufs)
 ```
 
 If `entry` is a directory or hamt shard, passing `offset` and/or `length` to `entry.content()` will limit the number of files returned from the directory.