Tentative fix to TEB convolution

[keskitalo]

Re-interpret the TEB convolution:

• T-component requires simple intensity-only convolution
• EB-convolution mixes the input Q/U sky components in harmonic domain by mapping

a^E_lm -> a^E_lm cos(2psi) + a^B_lm sin(2psi)

and

a^B_lm -> a^B_lm cos(2psi) - a^E_lm sin(2psi).

The cos(2psi) and sin(2psi) terms are evaluated separately and co-added. In the absence of a HWP, the sin(2psi) term is zero and the second polarized convolution can be omitted.

[keskitalo]

I did a full source format. that should make the merging easier.

[keskitalo]

Feel free to merge into your branch. It is probably time to request a review and merge to toast3