stream chunking fails for very large documents

[ehennum]

Reported in:

http://stackoverflow.com/questions/29691353/read-the-large-document-using-marklogic-nodejs-api

[ehennum]

A regression test could use a large document such as the concatenated plays of shakespeare (without the DTD and XML prolog and with a root element containing all of the PLAY elements).

[ayuwono]

Verified passed and added test: nodejs-documents-write-large.js