Reduce computation time

[jguarato]

This PR introduces two optimizations aimed at reducing computation time during modal analysis:

1. Precomputation of Partial Model Matrices
A portion of the model matrices (excluding contributions from bearing elements) is now precomputed during the instantiation of the Rotor object. Previously, this computation was repeated every time matrix build methods were called. This change avoids redundant calculations and improves efficiency.

2. Globalization of the circle Variable in the Orbit Class
The circle variable in the Orbit class has been refactored to a global constant. This eliminates the need to recalculate the CIRCLE array upon each instantiation of an Orbit object, further reducing unnecessary overhead.

The following test was performed by varying the values of n and freq:

rotor = rs.compressor_example()
rotor.run_campbell(speed_range=Q_(np.linspace(0, 3000, n), "RPM"), frequencies=freq)

The results are summarized in the table below. The optimized version achieved up to 37% reduction in execution time, with an average improvement of approximately 30% across different configurations.

Frequencies (freq)	Speed Range Size (n)	Time (Old version) [s]	Time (Modified) [s]	Difference (%)
9	30	5.29	4.12	22.24%
9	300	28.61	18.17	36.47%
9	900	86.79	54.32	37.41%
15	30	3.80	2.51	33.87%
15	300	35.23	23.09	34.47%
15	900	105.49	70.16	33.50%
21	30	4.77	3.29	30.95%
21	300	44.68	30.44	31.88%
21	900	132.94	92.59	30.36%