Implement a resource estimator tool

[arostamianfar]

It is currently not trivial to estimate the resources needed to run the pipeline (particularly amount of disk space). This is especially important when processing large (10TB+) datasets as cost becomes an important factor.

The idea is to write a tool that preprocesses the data (similar to the header merger pipeline) and outputs the minimum resource requirements for running the pipeline.

Note that estimating the disk size is not as trivial as summing up the size of all files being processed as there is an overhead for each record depending on the number of samples, info, and filter fields.

[arostamianfar]

We should extend the scope of this tool to provide warnings about incompatible header fields as well. See https://groups.google.com/d/msg/google-genomics-discuss/r4_Kt2MITng/mrKieJjDAQAJ