prepare in parallel after reading

[TingDaoK]

Issue #, if available:

• The only step for preparation needs to be serial is reading from the original body for now.
• So, invoke the callback after reading from separate threads to make the following steps to happen in parallel
• For s3 express, which we can hit 100Gb/s upload from a 30 GiB streaming data, the rest of the steps happens in the critical path make the performance drops to around 60 - 70 gbps
• With this change, we can hit 100gbps upload again with parts level checksum enabled for that workload.

Description of changes:

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.