Add a bidirectional coder for Protocol Buffers' Text Format

[joshsh]

For practical reasons, Hydra needs to integrate more deeply with Protocol Buffers. It has become common to define a model in Hydra (for the sake of a richer language with better support for transformations, domain-specific validation, etc.) and use Hydra's Protobuf coder to also materialize the model in Protobuf. However, passing data between code generated from the Protobuf, and code generated directly from the Hydra model, has necessitated hand-written mapping functions for each and every type. These functions are time-consuming to write, although tools like GitHub Copilot can cut down the tedium by at least half. We need a better way.

Once #47 is done, the better way will be to write a bidirectional coder to translate between Hydra typed term representations, and one of Protobuf's built-in serialization formats (an alternative would be to write code to generate Java mappers for a given model, or Go mappers, etc., but this would need to be done on a language-by-language basis due to differences in the Protobuf APIs in each language). The contenders are:

• Binary Format -- most compact and efficient, and universally supported
• JSON Format -- would be convenient due to existing support for JSON in Hydra. Unfortunately, there is uneven support for Protobuf JSON across language-specific implementations. E.g. support in Go is pretty poor.
• Text Format -- a human-readable format which is well supported in both the Java and Go implementations of Protobuf, and presumably in a number of other languages

I propose to start with the Text Format for the sake of ease of development, debuggability. We can proceed to the binary format if that goes well. Note that a new Hydra model for the Text Format will also be required.