fix(flagging): increase default threshold for static weight masking

[ljgray]

Since this is just supposed to catch numerical issues, I'm moving the default thresholds much closer to the max/min float32 values.

I'm also interested in feedback about making this baseline-independent and just spitting out a mask, since it's adding some baseline-dependent flagging at the moment.

[sjforeman]

The change in the default values seems fine, since the new defaults are used in the rev07 daily config anyway.

I don't have any opinions about making SanitizeWeights baseline-independent, since I don't know what fraction of the data the current SanitizeWeights typically catches, or whether the flagged samples are localized in frequency or time (vs. spread out more randomly, which would incur a larger loss of data by moving to a baseline-independent scheme). I'd be fine with punting that issue to the future...