I can’t spot any mistake. Please correct if I am wrong.

Thanks.

Thanks for spotting the mistake in the code. Corrected !!!

The Rayleigh channel should be as

h= sqrt(1/2)*((randn(1,1000) + 1i*randn(1,1000) ));

here N=1000 is not the taps as in case of frequency selective channel.
Its slow varying/quasi-static frequency flat channel with
1000 realizations of the single-tap channel over which the monte-carlo simulation has been performed.
Law of large number relates the time average and statistical average for large N.

Yes. The channel is assumed to be flat
Even for a quasi-static channel, the channel varies slowly over one block that it can be considered constant. Thus the above code is still valid

h= sqrt((randn(1,100).^2 + 1i*randn(1,100).^2 ));

Q(1) It shows channel is flat for 100 bits. Is it mean 100 taps?
and
Q(2) If channel is quasi static then what ‘h’ would be ?