refactor: match upstream grouping in FormatNumberForMobileDialing

[twcclegg]

What this is — and what it is not

No behavior change. This is a parity/robustness refactor, not a bug fix.

The divergence

Upstream Java (PhoneNumberUtil.java:1535-1551) guards the whole disjunction:

if ((regionCode.equals(REGION_CODE_FOR_NON_GEO_ENTITY)
     || ((regionCode.equals("MX") || regionCode.equals("CL")
         || regionCode.equals("UZ")) && isFixedLineOrMobile))
    && canBeInternationallyDialled(numberNoExt)) {

The port dropped the outer parentheses, so && binds tighter and reads as NON_GEO || (MX/CL/UZ && isFixedLineOrMobile && CanBeInternationallyDialled) — the dialling check applies only to the MX/CL/UZ arm.

Why it is not observable

CanBeInternationallyDialled returns true for every non-geographical number by construction, before it consults any metadata:

var regionCode = GetRegionCodeForNumber(number);
if (!IsValidRegionCode(regionCode))
    // Note numbers belonging to non-geographical entities (e.g. +800 numbers) are always
    // internationally diallable, and will be caught here.
    return true;

So in the only case where the grouping matters — regionCode == "001" — the guard is unconditionally true and both shapes reduce to the same thing. For MX/CL/UZ the two forms are already identical. The expressions agree on every reachable input; this is a structural invariant (any country calling code whose region list is ["001"]), not a fragile coincidence.

Tests

This PR adds UN001/InternationalTollFree assertions (the non-geo arm) and MXMobile1/MXNumber1 assertions (the MX arm) to TestFormatNumberForMobileDialing, closing a real gap — neither branch had any prior coverage.

These tests do not, and cannot, distinguish the old grouping from the new one — per the equivalence proven above, whenever the non-geo/A branch is true, CanBeInternationallyDialled is provably always true too, so no assertion exercising that branch can tell the two shapes apart. They're legitimate incidental coverage of previously-untested formatting paths, not regression tests for this specific change. The only argument for the fix itself is the algebraic one above.

Why land it anyway

The equivalence is a coincidence of an invariant living in a different method. If CanBeInternationallyDialled ever starts consulting non-geographical metadata — which upstream could do without it looking like a breaking change here — this port would silently diverge on formatting, with no test able to catch it. The port's value is being a faithful mirror of upstream, and matching the grouping costs two characters.

Verification

• dotnet build csharp/PhoneNumbers.slnx -p:TargetFrameworks=net10.0 — 0 warnings, 0 errors.
• dotnet test csharp/PhoneNumbers.slnx -p:TargetFrameworks=net10.0 — 440 + 51 passed, 0 failed, 0 skipped (unchanged counts, as expected for a behavior-preserving change).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.64%. Comparing base (8ad7fa2) to head (f017f09).
⚠️ Report is 5 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #463      +/-   ##
==========================================
+ Coverage   87.51%   87.64%   +0.12%     
==========================================
  Files          43       43              
  Lines        3886     3885       -1     
  Branches      991      991              
==========================================
+ Hits         3401     3405       +4     
+ Misses        280      277       -3     
+ Partials      205      203       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.