Uses a different serialization format

[nhatcher]

We currently use JSON and serde_json to serialize/deserialize models. Experiments show that other formats might be better suited for the task. For instance on a 7.5Mb Excel file:

Parse into json: 604.852904ms
Parse from json: 1.306669155s
Parse into binary: 112.957899ms
Parse from binary: 225.933871ms

See the remote branch for some experiments:

https://github.com/ironcalc/IronCalc/tree/experiment/nicolas-bincode

Just run cargo build --release and then:

./target/release/bench filename.xlsx

for a large workbook

Note that in a few places we will also need to use something else than JSON for serializing/deserializing, but that would remove the whole serde dependency.

It would be interestinmg to see the wasm file size after this.

See also:

https://github.com/djkoloski/rust_serialization_benchmark

[nhatcher]

FYI @fosdickio