Separate parsing from the parse results

[chafey]

Currently the parse results (e.g. codestream.segments, codestream.tiles) is tightly coupled to the parsing logic (codestream.decode()) logic. These should be separated for a few reasons:

1. the parsing code can be re-used without having to do decoding. A few examples come to mind - a simple codestream dump utility, code that needs to implement sub resolution fetching, JPIP server
2. The decoding logic can be re-used independently of the HTJ2K code stream itself. A few examples come to mind - JPIP client, resumable decoder (e.g. progressive decoding client side)

Some initial thoughts on how to proceed:

1. Pass a callback class to the parser which gets called as the parser works. This is like a SAX based parser. The callback should be able to cancel or pause parsing
2. A callback can be used to create a codestream object from the results of the parser

[chafey]

Discussed with pantelis, and we want to break up Codestream into three separate classes:

1. Data structure that holds the data in a codestream (with no parsing logic)
2. A parser that invokes a callback with the data it parses
3. A parser callback that populates 1 with the results of 2

[chafey]

Requirement: The callback should be able to pause the parsing. While paused, new datg