Improve Feedpositions metric: bad_feeds_percent

[nritsche]

@jrs65 and @ssiegelx suggested that known bad feeds (from flagging broker) should get excluded from this metric and from the pre-calculated STD, which should then be much lower.

[ssiegelx]

DocLib 781 contains an ipython notebook that explains how to query the flagging broker for the inputs that it identified as bad at any given time. @cahofer

The idea is to have the feed position analyzer query the flagging broker for the bad input list at the time of source transit. It would then ignore those inputs, and send an alert if there are any other inputs that have outlier feed positions. The chan_id of those inputs should be listed in the alert.

[cahofer]

I am setting the feed positions that were flagged by the flagging broker to nan. This way when we plot the data, those outliers are not shown.

Since the percentage of bad feeds will therefore be rundandant we could export the feed number not meeting a certain threshold like suggested in issue 116 @ssiegelx