Enforce some limits during view indexing

This is mostly motivated by CBSE-811.

During indexing, many non-optimal and bad things can happen due to the fact views are defined by user supplied JavaScript code and documents can be too large, taking a long for the JavaScript engine to parse the documents, which in turn steals a lot of time for the cluster manager operations.

Therefore add some configurable rules like:

1) Skip processing of documents larger than some threshold, and log a message when this happens;

2) Ensure for a single view and document (map function call), no more than a certain threshold of bytes are emitted (sum of byte size of all keys and values emitted);

3) If too many keys are emitted for a single document, fail early instead of during a later indexing operation that touches an updated version of the document (or deleted version).