[nabobalis]

Summary

 SourceSpectra.temperature_response(channel)
-  interpolate the spectra onto channel.wavelength
-  integrate over channel.wavelength
+  interpolate channel.wavelength_response() onto the spectra wavelengths, 0 outside
+  integrate over the spectra wavelengths

Interpolating the spectra onto the channel wavelengths drops or multiplies lines narrower than the channel sampling.

Evidence

• Before: in the new test_temperature_response_line, a 0.02 Å unit-integral line gives 50.5× the response at the line when it sits on a channel wavelength, and 0 when it sits between two.
  After: both give the response at the line. All response tests pass.
• Against SSW aia_get_response(/temperature,/dn) for the seven AIA EUV channels (aiapy's Channel from refactor channel on top of abstract channel class LM-SAL/aiapy#419, SolarSoft V9 emissivity), the worst relative difference is 1.1e-4 before and 3.7e-5 after. The remainder comes from SSW's hc = 12398 eV Å.

Merge Danger

Door: two-way

Blast Radius: values

Every temperature_response result changes: slightly for smooth spectra, and a lot for line spectra sampled more finely than the channel.