This PR includes three changes to make sure the methodology is consistent:
Correct the conversion rate of bytes to kb (was using the kib conversion) basically change 1024 to 1000 in all conversions
Calculate bytes_per_imp directly from bytes_per_session and imps_per_session for clarity (old method was confusing because it went all the way to bytes_per_session_sec just to divide by secs again, this is what had resulted in the bug we saw this morning)
Add embodied network emissions to the data transfer emissions for ad selection
This PR includes three changes to make sure the methodology is consistent: