The reason for choosing the symbols 0 and 7 (0 and 14 for 30kHz etc. ) is to synchronize the start of these symbols with time at 0, 0.5ms, 1ms, 1.5ms and so on. If you simply divide the available cyclic prefix to the number of symbols, the symbol at the middle will not start at multiple of 0.5ms. That’s why the first and middle symbol are given larger CP. Some of these symbols should have to receive larger CP so they are the chosen ones.
The frame structure is defined so that the receiver can synchronize in time with the transmitter. It’s easier if everyone knows that a frame has 10ms, a subframe 1ms and certain symbols start at multiple of 0.5ms