add warning around opening suspicious (non-.adam extension) ADAM files

[ryan-williams]

Porting this over from guacamole #266:

I just got bit by trying to open a path that was a directory (I meant to pass a BAM); failure mode was not the most clear:

Exception in thread "Driver" java.io.IOException: Could not read footer: java.lang.RuntimeException: hdfs://demeter-nn1.demeter.hpc.mssm.edu:8020/user/willir31/data/set3/normal/set3.normal.fq/part-01487 is not a Parquet file. expected magic number at tail [80, 65, 82, 49] but found [67, 68, 67, 10]
    at parquet.hadoop.ParquetFileReader.readAllFootersInParallel(ParquetFileReader.java:238)
    at parquet.hadoop.ParquetFileReader.readAllFootersInParallelUsingSummaryFiles(ParquetFileReader.java:179)
    at parquet.hadoop.ParquetInputFormat.getFooters(ParquetInputFormat.java:389)
    at parquet.hadoop.ParquetInputFormat.getFooters(ParquetInputFormat.java:361)
    at parquet.hadoop.ParquetInputFormat.getSplits(ParquetInputFormat.java:245)

From digging around, #190 seems to imply that there is a decent amount of convention behind assuming ADAM parquet files will have an .adam extension; any interest in:

• requiring presumed-ADAM files to have a .adam extension (throwing an Exception if the extension is not recognized)
• printing a warning
• Arun suggested: catching such an Exception and attempting to throw a new Exception with a better inferred message, e.g. inspecting whether the failing path is actually a directory and making this explicit in the re-thrown exception?

[fnothaft]

+1, I would be interested. I like the idea of moving to the conventions from #190, with the slight change of .align.adam to .read.adam.

[ryan-williams]

how about I just start by throwing an exception if we get to the "load as ADAM" code-path and it doesn't end in .adam?

[fnothaft]

Cool by me.