Improve method of determining lat, lon index from coordinates about rocketpy

[giovaniceotto]

Solving this was much harder than I thought it would be. But I have done it in the latest commit.

Turns out that some weather files present latitude values in in descending sorted order, such as ERA-Interim files for South America. This is the reason that the last approach was to check if the input coordinates are in between two consecutive coordinates of the weather file by using a sign multiplication method, fail-safe in terms of ascending or descending order.

However, it is true that it was slow. To make it faster, bisect was used and descending grids were dealt with by reversing them, than applying bisect and than reversing them again.

Furthermore, instead of operation on the masked arrays of latitudes and longitudes given by netCDF4, the arrays were first converted into simple lists. This may present problems if the grid has empty values, but this is not generally the case with the common weather models. Since it improves significantly the speed of the code, I decided that it is definitely worth it.

In the end, we achieved a speed increase varying from 10 times to 1000 times, depending on size of the latitudes and longitudes list, dropping from ms to micro seconds.