[core] Fix FormatTimeSys rendering the same timestamp inconsistently (#3225)

[steps-re]

Fixes #3225.

Bug

FormatTimeSys() renders a steady_clock timestamp as wall-clock time, but it combined whole-second wall time from ::time() with the sub-second part of steady_clock::now(). Those clocks have unrelated sub-second phases (steady counts from an arbitrary epoch), so near a second boundary the same timestamp rendered one second apart between calls — the ±1s discrepancy in the issue. Per-call re-sampling of two clocks non-atomically also adds ~1µs jitter.

Fix

Map the timestamp through a single steady↔wall reference pair, captured once at load time, in one consistent microsecond domain. A given timestamp now always formats to exactly the same string. The arithmetic is factored into a pure, testable overload FormatTimeSys(target_us, steady_now_us, wall_now_us). This touches only the [SYST] logging/trace formatter — not the transport path.

Tradeoff (for maintainer review)

This caches the steady↔wall offset for stability. If the wall clock is NTP-stepped mid-run, [SYST] times drift by the step amount but stay internally consistent. The alternative — live re-tracking with only the flicker removed — is also reasonable; happy to switch to that if you prefer. Flagging since it's your issue and touches a utility.

Verification

• New Sync.FormatTimeSysStable sweeps the reference across misaligned steady/wall second boundaries and asserts a fixed timestamp renders identically — fails on the old arithmetic, passes with the fix.
• Full unit suite: 271/271 pass. --gtest_repeat=200 on Sync.FormatTime* → 200/200 (also removes a latent flake in the existing back-to-back-equality test).
• Built with -DENABLE_UNITTESTS=ON, no new warnings. DCO sign-off included.

[ethouris]

That's a very good solution of the problem, but I don't like the way how the code is shaped.

If the goal of providing a second FormatTimeSys function was to allow exposure of the raw values for the sake of the UT, this should be done more direct way. Sure, there's no other (reasonable) way to expose this for the UT than to place it in the header file, but this should be definitely presented as an internal version, not intended for any other use. So, in short, the sync.h file should contain

std::string FormatTimeSys(const steady_clock::time_point& time);

// Exposed for testing purposes only
struct SysClockReference;
std::string FormatTimeSysInternal(const steady_clock::time_point& time, const SysClockReference& rf);

Then the normal FormatTimeSys function should then pass the parameters directly, while the value of SysClockReference should be created as a static local variable - this is the only acceptable form of a C++ singleton (safe against threads and global variable initialization ordering).

It might be also worth a shot to split the implementation of FormatTimeSysInternal into the part that does all these time value conversion in order to produce target_wall_us, and the other that formats this value into text.

[steps-re]

reshaped as you suggested. SysClockReference and FormatTimeSysInternal now sit in the header under an "exposed for testing purposes only" banner, and the public FormatTimeSys keeps its single signature with the reference as a static local. also took your optional suggestion and split the internal impl into the steady-to-wall conversion and the text formatting. full suite passes locally, 276/276, and the regression test now goes through FormatTimeSysInternal.

[steps-re]

@ethouris bump when you get a chance. the reshape from your last review went in on the 28th and CI is green. happy to keep iterating if the shape still isn't what you had in mind.

[ethouris]

Sorry, we are a bit in a hot water now, hence the delay.

Looks very good; if you can merge the latest master to it I'll be appreciated. I need to see @cl-ment to finally look into it, but I think it can be assigned now to the next release.

[steps-re]

merged latest master in, no conflicts, and it does not touch anything our diff does.

built locally with unit tests on, full suite is 210/210 green including Sync.FormatTimeSysStable. no rush on the review, whenever your side settles down.