Move spin calculations to compute_constants

[cemitch99]

This PR moves many computations for the spin push outside the particle loop into compute_constants, to improve performance.

• Quad (note that a typo was fixed in the k < 0 case)
• ChrQuad
• Sbend
• Sol
• Add benchmark test

Close #1288

[cemitch99]

This is a necessary change resulting from the typo fix in Quad.H for k < 0.

[cemitch99]

There should only be one factor of gamma_ref in line 345, which is fixed in the revised line below.

[ax3l]

@cemitch99 uh, I want to do a small PR for Python bindings and benchmarks first, so we can see first hand with the new CI #1309 how much this gains.

[ax3l]

Let us merge #1313 and then once it has run on development we rebase this PR to run it again and see the impact of the changes.

[codspeed-hq]

Merging this PR will improve performance by 31.53%

⚡ 3 improved benchmarks
✅ 33 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_Sol[spin]	1.2 s	1 s	+14.65%
⚡	test_Quad[spin]	848.8 ms	645.4 ms	+31.53%
⚡	test_Sbend[spin]	1,180.2 ms	914.6 ms	+29.04%

---

_{Comparing cemitch99:move_spin_constants (785a235) with development (3ace399)}

[ax3l]

I think this captured all elements we have with spin support so far. The only other 3 are the drift models, which don't affect spin.