*   Make the Bigtable client connection pool configurable & increase the
    default # of connections for performance.
*   Added derivative of `tf.random_gamma` with respect to the alpha parameter.
*   Added derivative of `tf.igamma(a, x)` and `tf.igammac(a, x)` with respect to
    a.
*   Modified Bessel functions of order zero and one.
*   Add FillTriangular Bijector to create triangular matrices.

* 🐛 Reenable `allow_arbitrary_types` when only 1 argument is used on the API endpoint. PR [#13694](https://github.com/fastapi/fastapi/pull/13694) by [@rmawatson](https://github.com/rmawatson).
* 🐛 Fix `inspect.getcoroutinefunction()` can break testing with `unittest.mock.patch()`. PR [#14022](https://github.com/fastapi/fastapi/pull/14022) by [@secrett2633](https://github.com/secrett2633).

### Refactors

//go:norace func (out *Nat) Mod(x *Nat, m *Modulus) *Nat { out.resetFor(m) // Working our way from the most significant to the least significant limb, // we can insert each limb at the least significant position, shifting all // previous limbs left by _W. This way each limb will get shifted by the // correct number of bits. We can insert at least N - 1 limbs without // overflowing m. After that, we need to reduce every time we shift. i := len(x.limbs) - 1 // For the first N - 1 limbs we can skip...

//go:norace func (out *Nat) Mod(x *Nat, m *Modulus) *Nat { out.resetFor(m) // Working our way from the most significant to the least significant limb, // we can insert each limb at the least significant position, shifting all // previous limbs left by _W. This way each limb will get shifted by the // correct number of bits. We can insert at least N - 1 limbs without // overflowing m. After that, we need to reduce every time we shift. i := len(x.limbs) - 1 // For the first N - 1 limbs we can skip...