Fix flaky CBOR test

[graebm]

Issue:

Rare random failure of "CborTimeStampTest" (see logs)

Diagnosis:

This test does:

• get timestamp from current `std::chrono::time_pointstd::chrono::system_clock
• CBOR encode timestamp as seconds-as-64bit-floating-point
• decode timestamp
• compare original timestamp with encoded/decoded timestam
  • deal with floating point precision differences by casting to milliseconds-as-integer before comparing

The flow above seems good. There are some differences in floating point precision, but we ultimately compare the values at lower precision (milliseconds-as-integer). So what's up???

Description of changes:

Fix helper function that loses lots of precision, due to too much casting back and forth between floating-point -> integer -> floating-point -> integer

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

[graebm]

we were casting from float->int->float here, now it's float->float->float

well ... some of those "float" are weird "std chrono ratio" things, but you know what I mean