Improve error handling in sdm.fit_ and sde.fit_ functions

[cwhanse]

Diode model and diode equation fitting can fail for a variety of reasons: bad input IV curve data, lack of a well-defined minimum in the objective function, poor initial guess.

Current behavior is to let the supporting package (scipy, pysam) raise and handle errors. This behavior gives the user little information about the failure, and causes pvlib execution to stop rather than move on to the next IV curve.

pvlib could

• provide information about the failure from the supporting package
• use try/except conditions to avoid stopping execution at the first problem IV curve

See discussion at #1226 #1228 #1014