Consider CQI as something like saying that for codeword 0 I am expecting to be able to decode transport block of size X at MCS Y within a bler of 10 percent.
Hence CQI is not reported for individual antennas as transport block is basically coded via the Codewords.
We have only 2 defined so CQI is reporter for max 2 CW no matter you use 4 layers or 8 in the downlink (Tm9 like BF).
No, if we just do an exact copy of the cw0 and cw1 and map it to 4 layers, then essentially we are using only 2 layers in a sort of transmit diversity arrangements.
For 4 layer transmission, we have 2 Codewords which should be mapped to 4 layers so we essentially take some data from each cw and map it to individual layers. This is why 4 layer transmission can increase the data rate.
The equations for this mapping is defined in 36.211 section 6.