dbSta: fix ObjectId collision between ConcreteCell and dbModule

[openroad-ci]

Fixes #9661

Summary

• Tag ConcreteCell/ConcretePort IDs with CONCRETE_OBJECT_ID (0xF) in dbNetwork::id() to prevent collisions with dbModule IDs (tag 0x8)
• Guard against top_cell_ in id(Cell*) — it is a ConcreteCell but isConcreteCell() returns false for it, causing a crash when reinterpret_cast is called to dbObject
• Add regression test TestWriteVerilog.RemoveCellsIdCollision (CMake + Bazel)

Problem

write_verilog -remove_cells silently drops hierarchical module instances due to an ObjectId collision.
CellSet uses CellIdLess (comparing by network_->id(cell)) for ordering.
Two different Cell* types produce overlapping IDs:

• ConcreteCell (liberty/LEF): raw sequential counter ccell->id_ (0, 1, 2, ...)
• dbModule (hierarchical): (db_id << 4) | 0x8 (24, 40, 56, ...)

When these IDs collide, CellSet::contains() treats a dbModule as equal to a ConcreteCell, causing writeChild() to silently drop hierarchical instances.

Solution

Introduce a new tag CONCRETE_OBJECT_ID for ConcreteCell and ConcretePort to avoid the ID collision.

Tag allocation

Tag	Value	Object Type
DBITERM_ID	0x0	dbITerm
DBBTERM_ID	0x1	dbBTerm
DBINST_ID	0x2	dbInst
DBNET_ID	0x3	dbNet
DBMODITERM_ID	0x4	dbModITerm
DBMODBTERM_ID	0x5	dbModBTerm
DBMODINST_ID	0x6	dbModInst
DBMODNET_ID	0x7	dbModNet
DBMODULE_ID	0x8	dbModule
CONCRETE_OBJECT_ID	0xF	ConcreteCell, ConcretePort (NEW)

🤖 Generated with Claude Code

[github-actions]

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

[gemini-code-assist]

Code Review

This pull request aims to resolve an ObjectId collision between ConcreteCell and dbModule that causes hierarchical instances to be dropped during Verilog generation, by introducing unique tags for ConcreteCell and ConcretePort IDs. While the fix for dbNetwork::id(const Cell* cell) is correct, the corresponding fix for dbNetwork::id(const Port* port) is incomplete. It fails to apply the necessary tagging logic to ConcretePort objects when hierarchy is enabled, allowing ID collisions to persist for ports and potentially leading to similar integrity issues in the generated output. Consistent application of the tagging logic across all concrete objects is required to ensure the robustness of the fix.

[jhkim-pii]

Checking ORFS QoR impact w/ The-OpenROAD-Project/OpenROAD-flow-scripts#3957.

[github-actions]

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

[maliberty]

I don't see an actual use of this bit? How is it checked?