Revert dev model components

During development of the 2018 JSON model, a variety of hacks were introduced to maintain consistency with the 2014/2018 source model. Most arise from rounding/precision issues in the legacy fortran codes and, to the extent possible, should be removed and changes documented with sensitivity tests.

• WUS fault widths in ModelLoader are computed using sin(δ * 3.14159 / 180°) and rounded
• CEUS faults rupture-sets have widths prescribed in properties elements as they were in the XML because widths from dips (e.g. 89°) and upper/lower depths are not identical
• Update CEUS vertical fault dips to 90° (there is another issue for this)
• Cheraw SSCn floating rupture has width corresponding to 50° dip
• MFD rates from slip rates on model load do not agree with rates from legacy codes; moment rates aren't equal due to differences in width length (probably can't fix this)
• Meers low rate out-of-cluster branch includes scale factor (2.0e-6 * 2.47207207207) to achieve Meers rate/hazard match
• Cheraw low rate out-of-cluster branch (1 eq) includes scale factor (3.2996e-6 * 1.00387925805) to achieve Cheraw rate/hazard match
• CEUS rate trees contain mixed data formats, some are in years and some are in 1/years. USGS tended to have year based rates (e.g. 500-yr, 1000-yr) but SSCn HID db had low precision annual rates that do not yield clean, integer-based recurrence intervals. Probably want to have high precision annual rates that when rounded(1/AR) yield an integer value that can be used as a label.
• Cheraw had slightly different traces defined for the SSCn model necessitating the addition of a temporary fault section Cheraw (SSCn, GR).geojson that should be removed.
• Low rate SSCn cluster-out branch of new Madrid scaled by 1.0074257425 to match target 0.000499 in xml.
• Low rate SSCn cluster-in branch of new Madrid scaled to 0.000619802 (down from 0.00062) to match weighted average poisson rate 0.002657 from HID database in xml.