Hydrophones fixes

[forrestv]

• allow running hydrophones script directly on bag for offline analysis/debugging
• don't bother trying to find start position of ping, instead just trust low-level driver's triggering
• use pearson correlation instead of mean absolute difference to find relative delays, which should be more robust
• intelligently compute search window based on hydrophone spacing rather than assuming they're within 1 wavelength of each other
• optionally generate plots for debugging

See https://pastebin.com/fUn2Um0d for the results of running on the bags at https://ci.mil.ufl.edu/labelme/bags/navigator/2018-10-07/; results look decent, except for one outlier (marked) that is due to the motors being turned back on during the recording.