Add new jo_fixed_cos func and correct tan division

[alfanhui]

Changelog:

• Previous method of obtaining cos from sin(rad - pi/2) causes discrepancy between sin and cos values when used together. (At least the 3rd, 4th, and 5th DP were incorrect for cos!)
• Tan sees greater accuracy since it used cos and corrected use of fixed division.

Description: Here is a few math calculations to prove new cos(45degree) function accuracy:

You can compare sin/cos/tan values with this table at the 45 degrees mark: https://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1011&context=mathclass

I noticed this cos issue as I am creating a right-angle turn movement along a grid. In this gif, you can see when I turn right (EAST), the grid starts trailing off the screen, and then the other gif is with the fix:

Original cos():

Fixed cos():

Interesting, sglCos is showing an inaccuracy as well.. i'll leave that for someone else to look into!

[ponut64]

I think this is a good change, if only for that it is directly using fixed-point maths, and is an alternative method to SGL sine and cosine.

That said, it is using a lot of divisions. Sinus and co-sinus are not used often, so that is okay. All the same, having a fast approximate version would be a good idea.

Also, a VERY important note: REMOVE the use of the toFIXED macro. This multiplies by a floating point. That is VERY bad for performance. When converting an integer to a fixed-point, shift left by 16 (<<16). To convert a fixed to an integer, shift right by 16 (>>16).