Make block size limit part of the BlockEncoder API

Currently BlockEncoder interface just encodes passed input into bytes

https://github.com/multiformats/js-multiformats/blob/58117f2c1a90934903651b456ec3a09552d942e7/src/codecs/interface.ts#L6-L10

Problem is:

1. If you go beyond block size limit, your block won't be bit-swappable.
2. You may not even know you've gone past block size limit.
3. Most people is not even aware of the block size limit.

Proposed solution:

I would like to propose amending our BlockEncoder interface as follows:

export interface BlockEncoder<Code extends number, T> {
  name: string
  code: Code
  /**
   * Encodes given data. If `buffer` is provided data is written into it and
   * a `Uint8Array` view of the written bytes is returned.  If `buffer` is not
   * passed new buffer is allocated with a `byteLength` corresponding to
   * default block size limit in IPFS. If encoded data does not fit the `buffer`
   * `RangeError` exception is thrown.
   */
  encode: (data: T, buffer?: Uint8Array) => ByteView<T>
}

Idea here is that:

1. User should be able to optionally pass in the buffer to write data into.
2. If data does not fit the block size limit encode fails.
3. User is still able to encode blocks larger than a block size limit by passing in larger buffer.

This will be non-breaking change at the API level, but it would be breaking in the sense that errors will occur if block is larger than a block size limit. Never the less it seems like a better default than silently letting things slip.