Make SDF Cache size tunable

[brentrobinson]

The SDF provider has an in-memory cache of inserted and updated features. The cache limit is 500 features. When the cache size is exceeded, the features are flushed to the SDF file.

For some data sets, a larger cache limit leads to better feature insertion performance, which can speed up the initial population of these SDF files. However, the amount of improvement varies greatly by data set.

This ticket proposes a couple of modifications to make the SDF cache size adjustable:

1. Add a new environment variable called SDF_MAXCACHESIZE. When set, the SDF provider sets the cache size to the number specified. If the variable is a non-positive number, not numeric, or not set then the cache size defaults to 500 features.

2. Add a new SDF Provider connection parameter, called MaxCacheSize, which behaves the same way as the above environment variable. This allows the tuning of the cache size for individual SDF connections. If the environment variable is also set, the connection parameter takes precedence.

The above provides the opportunity to try different cache sizes to arrive at the optimal sizes for particular data sets.