Cicada-mpc

* Added support for Numpy 2.0.
* Fully documented PRZSProtocol and cicada.arithmetic.Field, and added a glossary to the documentation.
* Significant changes and simplification of the cicada.transcript API.
* Added support for generating consistency verification code transcripts.
* cicada.arithmetic.Field.uniform() could generate values outside the field in some cases.
* Renamed cicada.logging to cicada.logger to avoid shadowing the builtin logging module.

* Added new API for generating transcripts - detailed logs for advanced debugging, network tracing, and zero-knowledge proofs using MPC-in-the-head.

* Moved the PRZS implementation into its own protocol object.
* Began running regression tests using Python 3.12.
* Minor documentation improvements.
* Restored missing regression tests, improved code coverage, and removed obsolete code.

* Our first stable release!
* Split encoders into separate field arithmetic and encoding objects.
* `share()` and `reveal()` methods automatically apply a default encoding with optional caller override.
* API explicitly distinguishes between encoding-agnostic and encoding-specific operations.
* Consistent support for private/private, public/private, and private/public arithmetic, based on argument types instead of separate methods.
* Added configurable logging for message transcripts.
* Added `cicada generate-shares` command for pedagogy.

* Reorganized the documentation and added instructions for building the docs locally.

* Renamed ActiveProtocol -> ActiveProtocolSuite.
* Renamed AdditiveProtocol -> AdditiveProtocolSuite.
* Renamed ShamirProtocol -> ShamirProtocolSuite.
* Renamed ShamirBasicProtocol -> ShamirBasicProtocolSuite.
* Protocol suite encoders have been removed from the public API.
* Added multiply(), divide(), and reshare() methods to all protocol suites.
* Significantly improved FixedFieldEncoder.uniform() performance.
* Switched to a faster dot-product implementation for ShamirProtocolSuite.
* Added documentation on active security.
* ActiveProtocolSuite.verify() didn't work with scalar secrets.
* Calculator service supports unix domain sockets.
* Improved test coverage.
* Switched from setuptools to flit for builds.