Scale factor is determined at high level in the BSON type 7 binary. When an encoder is appending doubles that need to compress we will start with scale factor 0 when starting a new Simple-8b block.

Every double that is appended need their minimal scale factor needed to be able to represent the double exactly as an integer. 

When a double is appended that needs a larger scale factor than previous doubles the encoder need to determine what can compress the most values between:

1. Finalize current Simple-8b block and start a new block for the value that needed a larger scale factor.
2. Re-scale previous values to the new (higher) scale factor and check if they still fit within a single Simple-8b block together with the new value.