Minimap2 aligned reads are broken in HaplotypeCaller due to being mis-identified as FlowBased reads and crashing.

[jamesemery]

It seems that a few users have complained when they try to use LongReads that their minimap2 bams are failing with the message:
org.broadinstitute.hellbender.exceptions.GATKException: Unable to trim uncertain bases without flow order information

It looks like in AssemblyBasedCallerUtils.java:147 we are calling FlowBasedReadUtils.isFlow(originalRead) on every read and the presence of the tp read tag in reads is considered sufficient to flag reads as being flow based which finally causes them to fail. This check was probably misguided, we should really be checking flow-based identity (at this stage anyway) from the readgroup in the header to prevent any spurious read tags from god knows what aligners don't cause problems like this again. Alternatively we should thread the "isFlowBased" check down into this part of the code so its opt-in to treat flow based reads specially when clipping here.

[ilyasoifer]

@jamesemery looks unfortunate. Will fix this of course.