Ignore null opcode bytes within a cgm page

[tmecklem]

This PR changes the cgm page decoding to ignore null opcode bytes throughout a cgm page.

Background: While decoding cgm pages and examining the output, I noticed a recurring pattern of two 0-value GlucoseSensor records followed by a DataEnd in some of the pages (See json output example below). It looks like some subset of pumps append two null bytes after a DataEnd (0x01) record. These two bytes are being identified as sgv records with a value of 0.

I made the change in this PR to ignore nulls within a page and ran a before/after diff of every glucose page in the repo I could find. In each diff, the two sgv records are no longer present and the corresponding DataEnd record has a timestamp 10 minutes earlier (since the two glucose records are no longer offsetting the DataEnd relative timestamp).

  {
    "name": "GlucoseSensorData", 
    "date_type": "prevTimestamp", 
    "_tell": 235, 
    "sgv": 0, 
    "date": "2015-04-29T07:43:00", 
    "packet_size": 0, 
    "op": 0
  }, 
  {
    "name": "GlucoseSensorData", 
    "date_type": "prevTimestamp", 
    "_tell": 234, 
    "sgv": 0, 
    "date": "2015-04-29T07:48:00", 
    "packet_size": 0, 
    "op": 0
  }, 
  {
    "packet_size": 0, 
    "name": "DataEnd", 
    "date": "2015-04-29T07:53:00", 
    "date_type": "none", 
    "_tell": 233, 
    "op": "0x01"
  }

[tmecklem]

For an example of a glucose page with the null bytes, see https://github.com/openaps/decocare/blob/master/analysis/cgm/bewest/ReadGlucoseHistory-page-82.data

[scottleibrand]

LGTM. We should also have someone else with MDT CGM test this out in real-world use before merging.

[bewest]

Excellent!

[bewest]

If the implementation is wrong somehow, due to an edge case, what might you expect to see?