Combine MockRandomPostingsFormat and RandomCodec [LUCENE-4457]

This is really confusing. I think MockRandomPF is mostly obselete, in that it writes a random seed file (.sd) so that its symmetrical when reading the index.

this is no longer necessary (codecs must write what they need). But i think it does some other cool things, like randomizes more params like skipintervals and so on.

we should make sure RandomCodec does all the evil things it does, delete MockRandomPostingsFormat, and rename RandomCodec to RandomPostingsFormat.

Then add a real Random codec thats really a frankenstein and does things like switch up stored fields impl (#5298) and so on.

---
Migrated from [LUCENE-4457](https://issues.apache.org/jira/browse/LUCENE-4457) by Robert Muir (@rmuir), updated Oct 04 2012

The issue names MockRandomPostingsFormat and RandomCodec but no files, tests, or entry points. Start by locating both implementations and their codec-related tests, then determine whether RandomCodec covers the existing randomized behavior and what the separate frankenstein codec must include before removing and renaming components.

Written by the indexing model from the issue text.