BTreeCounters does not track bytes released.

[thompsonbry]

The BTreeCounters class tracks the #of bytes written on each index, but it does not track the #of bytes associated with records which have been deleted. Thus, it is not possible to compute the #of bytes for only the current records on the B+Tree. This means that estimates of the average page size based on the current #of nodes and leaves and bytes written are incorrect, at least for the RWStore.

Add a counter for bytesReleased and update logic which is using bytesWritten under the assumption that it reflects the WORM architecture, which does not recycle, and the scale-out architecture, which recycles by deleting entire files.

[thompsonbry]

There is also interest in:

- Making the counters restart safe (save them with the B+Tree checkpoint).

- Tracking the eviction rate for the index.

- Using the load size of the index nodes and leaves (from their persistent addresses) and possibly metrics about eviction rates to dynamically control the size of the write retention queue.

[thompsonbry]

BTree#removeAll() has an optimization for the RWStore. The performance counters need to be updated in that optimized code path as well.

[thompsonbry]

Once we have restart safe performance counters for the B+Tree we can rapidly report on the statistics for the different indices. DumpJournal? and DumpIndex? currently coordinate to do this, but this process involves a page scan of the index. With restart safe performance counters which correctly track the bytes utilized for nodes and leaves we could produce the same table "instantly".

[thompsonbry]

Change sets for RWStore history and persistent memory leaks following code review with MCutcher.

Committed Revision r5809 (against 1.0.x maintenance branch and TERMS_REFACTOR_BRANCH)

@see https://sourceforge.net/apps/trac/bigdata/ticket/217 (BTreeCounters does not track bytes released)
@see https://sourceforge.net/apps/trac/bigdata/ticket/329 (B+Tree should delete bloom filter when it is disabled)
@see https://sourceforge.net/apps/trac/bigdata/ticket/372 (RWStore does not prune commit record index).
@see https://sourceforge.net/apps/trac/bigdata/ticket/375 (Persistent memory leaks (RWStore/DISK))