Add check that data size is greater than topTreeSize

[aaronquantexa]

CL:

• Add check that topTreeSize parameter is greater than the input data size. Throws an exceptions stating this.

Motivation behind the change was on this raised issue: #21 about difficulty debug.

Before this change, the error would be:

requirement failed: Sampling fraction (1.002267573696145) must be on interval [0, 1]
java.lang.IllegalArgumentException: requirement failed: Sampling fraction (1.002267573696145) must be on interval [0, 1]

whereas now the happens earlier before doing any data transformations, and says:

org.apache.spark.SparkException: Invalid top tree size relative to size of data. Data to fit of size 441 was less than topTreeSize 442

In a previous commit, included a test for this which matched the string in the error message, but didn't see it as necessary.