The calculation actually still proceeds incorrectly,
but it is a matter of physics, not of programming,
and will be solved in next commits.

Like almost everywhere else, there is no need to distinguish between mass of KL and KS in the prefactors in amplitudes_LD.

Originally, the function `amplitudes_LD(par, K, l)` would send both long-distance coefficients - for KL and KS, despite the fact that it takes `K` as an argument; `amplitudes_eff` then always thrown one of them away.
The new implementation should be more transparent.

* The original `amplitudes` function renamed to `amplitudes_weak_eigst`
 as it actually yields amplitudes for K0, K0bar, not KL or KS.

* A new `amplitudes` function, which newly takes also `K` (to be `'KL'` or `'KS'`)
 as an argument, has been implemented.
 It makes proper linear combinations of amplitudes for K0 and K0bar
 and changes by sqrt(2) the normalization of the amplitudes
 (cf. arXiv:1602.00881 with 1711.11030).

* For l1=l2, `amplitudes` yields the imag/real parts of the S, P weak_eigs coefficients, as it should.
 Thus, taking real/imag parts is no longer tackled in `amplitudes_eff`,
 which now serves merely to add up short- and long-distance contributions, as it should.

* Relative signs between SD and LD contributions now tackled in `amplitudes_eff`.
 It is in accordance with the original version of this file;
 looks differently than in arXiv:1711.11030, but it is due to different WC
 conventions (see my notes for details).

As long as we neglect indirect CPV in kaons (which we do in kll.py),
WET predicts BR(KL,KS->e+mu-) = BR(KL,KS->mu+e-), so we do don't
calculate twice anymore at `br_kll_fct_lsum`.