Originally implemented as a workaround to the bson crate cloning every value it
comes across and it's looking like it shows significant improvement across the board
for serialisation (~80% improvement).
``` mongodb's bson time: [1.1160 us 1.1171 us 1.1183 us] Found 2 outliers among 100 measurements (2.00%) 2 (2.00%) high mild
serde_bson time: [201.99 ns 202.17 ns 202.38 ns]
Found 10 outliers among 100 measurements (10.00%)
4 (4.00%) low mild
4 (4.00%) high mild
2 (2.00%) high severe
```
There's a few pieces missing such as arrays and nested documents but they're not too difficult to add, it's just that it's 2:38am and I've smashed this out in an hour.
Pull requests welcome as always.