Improve performance of slide histogram

[slfritchie]

This is a patch by @jtuple ... I'm just creating the pull request to get the review & merge show on the road....

Partition slide histogram reservoir by erlang:system_info(scheduler_id),
as previously done with counter and spiral metrics.

[slfritchie]

+1 to merge.

Verification, using Erlang/OTP R15B02 or R15B03:

1. git clone git://github.com/basho/basho_bench.git
2. cd basho_bench
3. git checkout origin/slf-null-performance-enhancements
4. make
5. (cd deps/folsom ; git checkout 4b72c41) ... this is the basho/folsom repo head branch's HEAD commit
6. make
7. ./basho_bench /path/to/preload.config
8. tail -1 tests/current/put_latencies.csv | awk -F, '{print $1}' ... this is the total runtime, in seconds.
9. (cd deps/folsom ; git checkout jdb-faster-slide)
10. make
11. ./basho_bench /path/to/preload.config
12. tail -1 tests/current/put_latencies.csv | awk -F, '{print $1}' ... this is the total runtime, in seconds.

... where the preload.config file is this:

{mode, max}.
{concurrent, 100}.
{mode, max}.
{concurrent, 1}.
{duration, 3}.
{report_interval, 1}.

{driver, basho_bench_driver_null}.
{key_generator, {partitioned_sequential_int, 10000000}}.
{disable_sequential_int_progress_report, true}.
{value_generator, {fixed_bin, 1000, 42}}.
{operations, [ {put, 1} ]}.

On my MacBook Pro, I see the following runtimes omitted for steps 8 and steps 12:

• step 8: 107.713825
• step 12: 24.7412

That's about 4.3x faster, yay.