Refactoring exact norm

[RonanLegin]

In pspecdata.py, in the definition of q_hat (line 956), there is a for loop present (line 1052) in which within this loop, we calculate Q as Q = del_tau * self.get_Q(i, pol). This for loop is going over each delay mode "i" in which get_Q takes as a parameter. If we go to the function get_Q (around line 1557), it computes the exponential term and the integral of the primary beam. However, the integral does not depend on the delay mode. So back in the q_hat function, the integral is being calculated in the loop for every delay mode when in fact its result is constant. To change this, I created a separate function in pspecdata.py called get_integral_beam where the code that was in get_Q to compute the integral has been transferred to this new function. Outside of the for loop, I call get_integral_beam and store its result in a variable called integral_beam and then multiply it inside the loop as Q = del_tau * self.get_Q(i, pol) * integral_beam. This way we are always using the same result for each delay mode. This makes generating the power spectrum faster using exact_norm = True as a parameter in pspec. This change does not affect the time to generate the power spectrum when exact_norm is set to "False", as it doesn't use this for loop in q_hat. I also wrote a test function called test_get_integral_beam in test_pspecdata.py for my new function get_integral_beam.

[coveralls]

Coverage increased (+0.002%) to 96.819% when pulling 69c0d78 on refactoring_exact_norm into 7708f92 on master.

[saurabh-astro]

Looks great. I have made minor suggestions related to docstring.

[philbull]

I agree with @saurabh-astro's comments -- this is ready to go once those docstrings and comments have been updated.

[philbull]

There was a merge conflict, so I fixed that and did a bit of tidy-up. @Ronan1879 don't forget to pull before making any more changes!

[philbull]

Looks good!