add t_ccd_ref_alpha

[javierggt]

@jeanconn

This is the change I propose in #172 (this can just be merged there).

[taldcroft]

@javierggt - See the functional testing of this in https://github.com/sot/acdc/pull/87.

The upshot is that varying t_ccd_ref_alpha by +/- 10 C makes the trending plots noticeably worse or unphysical.

Full disclosure is that I have not reviewed the details of the model derivation. That said, my understanding was that the best-fit parameters a, b, c, d, e would be dependent on t_ccd_ref_alpha. I think this assumption is validated with the functional testing which demonstrates that the current hardcoded a, b, c, d, e give optimal results only for t_ccd_ref_alpha=265.15. Any variations in that parameter without corresponding changes in the other parameters will degrade the dark current scaling model.

Is that correct?