[x/programs] serialize/deserialize using serde

[samliok]

Serialization using Serde

Why?

In order to support more complex data structures moving across the WASM boundary we would need to define our own ways to convert them to/from bytes. This is what ProgramValues were used for. ProgramValue's implemented the Store trait with methods as_bytes, from_bytes & as_tag. Crates like Serde & serde_bare manage & implement these methods for us, and for more complex structs. This PR uses the from_slice and to_vec functions from the serde_bare crate to serialize/deserialize objects.

Advantages

• Developers are now limited to Serde's serialization types which are much more expansive than ProgramValues
• Program UX is greatly simplified. No need to convert to/from ProgramValues, can just use native types/user defined structs. See before and after at the bottom.
• Don't have to program serialization/deserialization which can be a non-trivial task.
• Doesn't increase binary size that much. TokenProgram went from ~228kb to ~222kb.

Disadvantages

• At least for now, we still need to define our own serialization for parameters. This includes any type that implements the Argument trait. So far these are ProgramContext, i64, Bytes32, and Address. The reason for this is b/c when we invoke an external program, we need to serialize all params into one byte array. Then in go we need to deserialize in order to pass on the parameters to the desired external function.
• Performance costs see @hexfusion's comment see performance section

Changes

• Remove ProgramValue, Store and related stuff
• Uses the from_slice and to_vec functions from the serde_bare crate to serialize/deserialize objects.
• Get and Store functions take in a param that implements Serialize or DeserializedOwned
• Argument trait, used during program_invoke
• Added Pokemon example to show using serde for custom structs
• Update all program examples
• Changes ProgramContext -> Context
• Added go tests for Pokemon and Lottery examples.

Performance

goos: darwin
goarch: arm64
pkg: github.com/ava-labs/hypersdk/x/programs/examples
                                        │   old.txt   │              new.txt               │
                                        │   sec/op    │   sec/op     vs base               │
TokenProgram/benchmark_token_program-10   8.948m ± 4%   9.777m ± 3%  +9.27% (p=0.000 n=10)

                                        │   old.txt    │               new.txt               │
                                        │     B/op     │     B/op      vs base               │
TokenProgram/benchmark_token_program-10   4.809Mi ± 0%   4.737Mi ± 0%  -1.50% (p=0.000 n=10)

                                        │   old.txt   │              new.txt               │
                                        │  allocs/op  │  allocs/op   vs base               │
TokenProgram/benchmark_token_program-10   43.32k ± 0%   46.83k ± 0%  +8.09% (p=0.000 n=10)

Compiling Rust using
./scripts/build.sh

Running test with
go test -v -timeout 30s -run ^TestTokenProgram$ github.com/ava-labs/hypersdk/x/programs/examples --count 1

Before

After

[samliok]

comments on bug

[hexfusion]

The perf cost of this in non trivial

goos: linux
goarch: amd64
pkg: github.com/ava-labs/hypersdk/x/programs/examples
cpu: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
                                        │   old.txt   │               new.txt                │
                                        │   sec/op    │   sec/op     vs base                 │
TokenProgram/benchmark_token_program-12   16.62m ± 4%   37.37m ± 2%  +124.88% (p=0.000 n=10)

                                        │   old.txt    │               new.txt                │
                                        │     B/op     │     B/op      vs base                │
TokenProgram/benchmark_token_program-12   4.811Mi ± 0%   7.611Mi ± 0%  +58.20% (p=0.000 n=10)

                                        │   old.txt   │                new.txt                │
                                        │  allocs/op  │  allocs/op    vs base                 │
TokenProgram/benchmark_token_program-12   43.33k ± 0%   116.28k ± 0%  +168.35% (p=0.000 n=10)

[richardpringle]

Minor comments, feel free to open an issue to fix later if you want to get this PR landed

[richardpringle]

LGTM

[patrick-ogrady]

Can we change this to init?