GRT: Fix FastRoute blocked tracks computation

[jfgava]

Summary

The problem. initBlockedIntervals converts blocked metal into a track count so it can reduce a gcell edge's capacity. horizontal_blocked_intervals_ / vertical_blocked_intervals_ are interval_set<int> — sets of disjoint intervals — and the old code rounded up each interval separately:

for (const auto& interval_it : intervals) {
  reduce += std::ceil(len(interval_it) / track_space[layer - 1]);
}

Every fragment therefore paid a full track's round-up, so the charge depended on how the blocked region happened to be fragmented, not on how much of it was covered. That makes it non-monotonic: deleting metal that bridged two fragments splits the span, both halves round up independently, and capacity drops even though less metal is blocking.

The fix. Sum the covered length across all intervals first, then round to tracks once:

int64_t blocked_length = 0;
for (const auto& interval_it : intervals) {
  blocked_length += std::abs(interval_it.upper() - interval_it.lower());
}
return std::ceil(static_cast<double>(blocked_length) / track_space);

The charge is now a function of coverage alone, so it can never rise when metal is removed. Extracted into blockedTrackCount() and used by both the vertical and horizontal loops — the defect was symmetric and both are fixed. int64_t accumulation avoids overflow on wide spans.

Scope. Removes the fragmentation artifact; still position-blind (it approximates blocked tracks as length/pitch rather than testing real track coordinates). Exact track counting remains the optional follow-up, which would need the per-layer track origin plumbed into FastRouteCore.

Type of Change

• Bug fix

Impact

Properly adjust g-cell capacity. This mainly impacts Metal 1 resources.

Verification

• I have verified that the local build succeeds (./etc/Build.sh).
• I have run the relevant tests and they pass.
• My code follows the repository's formatting guidelines.
• I have included tests to prevent regressions.
• I have signed my commits (DCO).

Related Issues

#10273

[gemini-code-assist]

Code Review

This pull request introduces a new helper function blockedTrackCount in FastRoute.cpp to calculate the number of blocked tracks by summing the covered length of disjoint intervals and rounding once, which prevents over-charging and non-monotonic behavior. It also updates several test outputs and configuration files to reflect these routing changes. The review feedback identifies a critical issue where a non-positive track_space could lead to a division-by-zero error in the new helper function, and suggests a safe guard check to prevent potential crashes.

[jfgava]

I'm running a CI. I'll need to update some metrics. But it seems ok