pdn: only consider routing layer obstructions for grid obstructions

[gadfort]

Changes:

• only consider routing layer obstructions from macro obstructions during via generation.
• Since the followpins are allowed to ignore the grid obstructions without this, the vias cannot be built in M1/M2 via arrays if the macro contains via blockages.
  This should largely be a no-op for most PDKs.

[gemini-code-assist]

Code Review

The pull request correctly modifies the via generation logic to ignore grid obstructions on non-routing layers (cut layers). This change allows via generation in areas where macros have via blockages, addressing the issue where followpins could ignore these blockages but the vias connecting to them could not. The logic is sound and includes a performance improvement by adding a break once a via is found to be obstructed. However, the new integration test macros_cells_halo_blocking is missing its golden DEF file (.defok), which will cause CI failures. Additionally, the BUILD file needs to be updated to include this golden file in the test's resource map.

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"