Improvements to hypergeometric motives

[kedlaya]

In the wake of #23671, a meta-ticket to record some proposed improvements to hypergeometric motives. Please add suggestions to this list, and make individual tickets for individual changes.

• Euler factors at tame and wild primes.
• Tune the p-adic precision in padic_H_value.
• Speed up the computation of p-adic Gamma and/or modify the computation of Gauss sums to eliminate redundant evaluations of p-adic Gamma. (See #24433 for a relevant optimization: caching coefficients of the Dwork exponential series.)
• Recode the hypergeometric trace formula in Cython.
• Speed up possible_hypergeometric_data for fixed weight. (See #23953 for one step: returning an iterator instead of a list.)
• Add the computation of the Frobenius structure on the hypergeometric equation, which can be more efficient than the trace formula when the degree is large.
• Refactor: should this really be under modular?