bulk-compress blocktree term suffix lengths [LUCENE-4179]

these should be fairly tiny numbers, it seems wasteful to use a byte-per-term when they typically only need a few bits.

also if we have them in bulk, its a step towards random access.

---
Migrated from [LUCENE-4179](https://issues.apache.org/jira/browse/LUCENE-4179) by Robert Muir (@rmuir)
Attachments: [LUCENE-4179_suffixBlock.patch](https://apache.github.io/lucene-jira-archive/attachments/LUCENE-4179/LUCENE-4179_suffixBlock.patch), [LUCENE-4179.patch](https://apache.github.io/lucene-jira-archive/attachments/LUCENE-4179/LUCENE-4179.patch)

Start by reviewing LUCENE-4179_suffixBlock.patch and LUCENE-4179.patch, then locate the blocktree term suffix-length handling they modify. Determine how the suffix lengths can be bulk-compressed and how that supports random access; done means the representation no longer wastes a byte per term while preserving the blocktree behavior.

Written by the indexing model from the issue text.