Need to handle poor height-width fits better

[mikedurand]

Sometimes the height-width fit functions fail. Need to add a checker for poor fits, and encode a constant width fix for these occurrences.

[mikedurand]

In the latest run on Version D data (by Ted, end of February 2026), for example here are some reaches with messed up fits:

• 73150600151: h_break has a value less than the WSE valid min; also h_break values are non-monotonic increasing
• 84207600171: h_break has nan values

[mikedurand]

The fix should simply set all height-width fits to constant values, in values written to timeseries, and set area to corresponding values. One way to go would be to set breakpoints equal spaced between valid_min and valid_max, all slope values to zero, and all intercept values to a constant width. I