Fix IEC 61850 millisecond timestamp rounding

[masarray]

Customer issue

Relay and ABB SAS can show an IEC 61850 event as 31.201 while ARSAS showed 31.200 for the same source timestamp.

Root cause

ARSAS formatted sub-millisecond DateTime / DateTimeOffset values directly with .fff. That exposes the millisecond component without first rounding the full timestamp. A source value such as 31.2006000 therefore rendered as 31.200.

Fix

• introduce one IEC 61850 nearest-millisecond presentation helper
• midpoint (.0005000) rounds forward
• correctly carry across second/minute boundaries
• preserve the original full-resolution timestamp in evidence; rounding is presentation-only
• use the same rule for FAT live/current IED time and ON/OFF evidence
• align FAT Excel export, standard PDF report, executive PDF report, and HTML evidence output
• use the same rule in the general IEC 61850 value formatter
• add permanent regression coverage for 31.2006 -> 31.201, midpoint behavior, second/minute carry, and raw evidence preservation

Scope

The patch intentionally does not change IEC 61850 acquisition, report timestamps at source, NTP/SNTP behavior, GOOSE capture clocks, or stored evidence precision. It corrects only millisecond-resolution presentation.

Validation

Final head 803179408d9052f7b53daeb14fcea292e9429c83 passed:

• Validate IO List Testing Release ARSAS 1.6.25 #154 — SUCCESS
• Validate SV evidence bundles Reduce P1 workspace noise and progressive GOOSE disclosure #205 — SUCCESS
• Build ARSAS #1012 — source/license invariants, solution build, full regression suite, portable publish and portable smoke test — SUCCESS

The earlier bot-generated report-alignment commit produced action_required checks with no jobs because GitHub does not recursively trigger normal PR workflows from the bot push. A final human-authored no-logic comment update retriggered CI on the exact final source above.