Replace most uses of quanta::Instant with std::time::Instant to increase the accuracy of time measurements

[tatsuya6502]

This PR does the following:

• Replace most uses of quanta::Instant with std::time::Instant to increase the accuracy of time measurements.
  • When the quanta feature of moka is disabled, std::time::Instant is used for all time measurements.
• Refactor time-related code to reduce memory overhead when using std::time::Instant.
  • Now the memory overhead of using std::time::Instant is the same as using quanta::Instant.
  • moka keeps three timestamps for each cache entry. (last accessed time, last modified time, and expiration time)
  • Now each timestamp is stored as AtomicU64 (8 bytes) in nanoseconds.
  • Before, each timestamp was stored as Mutex<std::time::Instant>, which is 24 bytes on Rust 1.83 for Linux x86_64 target.
• To avoid unexpected panics on some platforms, use saturated arithmetic for time calculations whenever possible.

quanta is a high-performance time measurement library and will be great for some use cases such as collecting performance metrics. However, it is not very suitable for tracking expiration times of cache entries because the time measurement may drift over time, and it may not be monotonic on some specific hardware (#472). Therefore, using std::time::Instant will be more appropriate for tracking expiration times of cache entries.

(I created PR #482 to disable the quanta feature of moka by default)

[codecov]

Codecov Report

Attention: Patch coverage is 96.91011% with 11 lines in your changes missing coverage. Please review.

Project coverage is 94.11%. Comparing base (dd17cc1) to head (5d03e02).
Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #481      +/-   ##
==========================================
- Coverage   94.18%   94.11%   -0.07%     
==========================================
  Files          44       43       -1     
  Lines       20558    20486      -72     
==========================================
- Hits        19362    19281      -81     
- Misses       1196     1205       +9     

[tatsuya6502]

I ran short benchmarks using mokabench on moka@v0.12.9 and one on disable-quanta-by-default branch. I ran them on my home computers; Linux x86_64 (WSL2) and macOS arm64. I do not see any meaningful performance difference between moka@v0.12.9 and one on disable-quanta-by-default branch. (I only used the default features of moka, so quanta was disabled on the benchmark using disable-quanta-by-default branch)

[tatsuya6502]

Merging.