Allow parameter bounds

[markcampanelli]

Once scipy's ODR algorithm is upgraded to support fit-parameter bounds, PVfit should support "arbitrary" parameter bounds in addition to parameter fixing. Each bound would likely have a default range that physically makes sense, but could accommodate broader user choices (if computationally compatible) to allow more exploratory work, e.g., allowing a negative shunt conductance to gauge the degree model discrepancy.

Should this work replace fixed-parameter booleans with single-point boundary intervals? Also, boundary endpoint inclusion/exclusion is an important consideration.

This issue was inspired by a question from Josh Stein at the PVfit tutorial at PVPMC2024.