New main survey minimum exposure efftime_etc

[akremin]

This addresses email thread [desi-survey 3581] initiated by David Schlegel, where he advocated for more stringent efftime_etc cuts on the data.

The code changes are minimal. We move from a cut at 5% of goaltime to fixed cuts of:

• dark: efftime_etc < 100s
• bright: efftime_etc < 20s
• backup: efftime_etc < 0.5s,

where the exposure is not flux calibrated or fit for redshifts if it fails the relevant criterion.

Below I include plots David provided for bright and dark time respectively showing all main survey exposures currently marked as good, and highlighting those that would now be cut in red:

The plots are in efftime_spec, but he verified that the cuts do not change with efftime_etc. The red points will be manually rejected to make past data consistent with future data.

[akremin]

I also improved some of the logging and comments to make things more readable by reducing the decimal precision of floats being printed. That unfortunately makes the relevant changes less obvious, but I felt the updated logging was useful while I was editing this function.

I tested this on two nights with exposures known to fail the new cuts: 20211114 and 20220120. Those exposures were all flagged under the new version of the code as being exposures that shouldn't be flux calibrated.