Zebra: Add `secp256k1` and possibly wrap its types.

[dconnolly]

The secp256k1 crate is a Rust wrapper around the well-vetted and constant-time C library that is actively maintained.

While including a non-trivial C dependency into our (so far) all-Rust project isn't the ideal answer, this library is a better choice in terms of maintainability (and offloading that to someone else) for us than some of the all-Rust alternatives out there, which are either less-great from a constant-time, implementation perspective or are just less mature than bitcoin-core/secp256k1.

[hdevalence]

Changed title so that we have one issue for integrating the dependency and one issue for adding transparent keys and addresses.

[dconnolly]

Pulling this in exposes the signing operations needed to verify script ops for transparent transactions without joinsplits.

[hdevalence]

The library requires building a context object and using the context, which is sort of awkward. Maybe the easiest thing to do would be to use lazy_static to build the context and store it as a "constant" somewhere in zebra-chain.

[saleemrashid]

Not to co-opt this issue, but it could be worth looking into sandboxing. As far as I know, libsecp256k1 makes no syscalls, so you could use the strictest sandboxing policy. I think Google has some excellent open source, cross-platform sandboxing libraries, but they're for C++ (could be useful for seeing which APIs to use on each platform).

[hdevalence]

Sandboxing seems great, but unfortunately I don't think that we have resources to do it in the medium term. I don't feel great about C libraries in general, but I feel that of all C libraries that a Rust project might use, this one is relatively less risk than others.