spec: clarify expectations and reduce recommended size of volume ID

[jdef]

Volume IDs uniquely identify a volume. They should generally be human readable and useful for logging (debugging and tracing) purposes. The spec defines the type of id as string, which is fine. There's some debate around what the size of an ID should be.

This issue proposes that the CSI spec recommends a size limit of the protobuf-encoded volume id field to 256 bytes. Such a size provides ample space for uniqueness and also represents the upper length limit of a token that's useful w/ respect to logging and readability.

[jdef]

PR #123 defines the length limit of an encoded volume ID as 1MiB, which I'd argue is MUCH too large.