After the move to 64-bit upload/download byte counts, these integers are different width than 'long' on 32-bit builds (e.g. Windows), so these numbers were wrong on all but 64-bit *nix builds.

The change here also uses the cool byte_estimate_to_str() function so that large byte totals are not expressed in 'K' but rather 'M', 'G', 'T' etc. suffixes.

The formatting is otherwise consistent with the pre-existing log line format save for the use of an uppercase suffix now.

Before:
@- 02:19p  T: 12   R:  0   P:  0   E:  0   F:  0   U:  0k 0   D:420k 1
After:
@- 02:07p  T:  0   R:  0   P:  0   E:  0   F:  0   U:  0K 0   D:420K 1

Before:
@- 02:23p  T:  0   R:  0   P:  0   E:  0   F:  0   U:  0k 0   D:1359k 1
After:
@- 02:26p  T:  1   R:  0   P:  0   E:  0   F:  0   U:  0K 0   D:1.3M 1

This fixes issue #519. I'm glad to hear that sysops still use these log files!