Create `labelReplacementCanonize` primitive

[dlongley]

Make it take a function that will be passed canonicalIdMap and return a new canonical bnode ID mapping to be used to generate canonical N-Quads from a canonicalized dataset.

Change hmacIdCanonize() to take an hmac and return a function that will create a map with keys that are the original bnode IDs and values that are the associated u<hmac digest> values. Alternatively, make hmacIdCanonize do that internally. The idea here is to better match the spec text.

Implement labelMapCanonize() to take labelMap and return a function that returns the reverse of labelMap.