SST read optimization for levels greater than 0

[justinethier]

For level 1 and greater, if key is in the bloom filter we can check that the key is between the file’s indexes. If not we have a false positive and can continue to the next file.

In addition to making the fix it would be helpful to setup a benchmark for this to measure any speed increase. Perhaps we can pair it with a test or utility function that generates a large dataset with many SST's above level zero.