Add Viper compatibility and via import

[supertjhok]

Summary of fixes

1) Config file: accept -c <chipsmith_config.toml> and improve errors

Files: src/main.cpp (or wherever main lives), src/chipsmith/chipFill.cpp

• Enabled the banner (welcomeScreen()).
• Hard‑coded config.toml removed. We now parse an optional -c/--config CLI flag and pass that filename through the tool flow. If the flag is omitted we fall back to config.toml for backwards compatibility.
• Added strict/clear error messages when required top‑level tables are missing from the config, printing the available keys before exiting.

Why: The ViPeR flow provides a chipsmith_config.toml separate from the top‑level config.toml. The original chipSmith always tried to read config.toml, producing confusing errors.

---

2) Read ViPeR’s new TOML keys and translate to chipSmith’s internal structures

File: src/chipsmith/chipFill.cpp

• We still read the original chipSmith tables:

  • [GDS_CELL_LOCATIONS]
  • [GDS_MAIN_STR_NAME]
  • [GDS_LOCATIONS]

• New: we now parse [Biasing_Coordinate.*] (ViPeR style) and translate it into the legacy GateBiasCorX map that placeBias() uses.

  • For each block (e.g. [Biasing_Coordinate.AND2T]), we take the first element of top = [<x>] as the bias x‑offset (in microns).
  • We map the logical gate name (AND2T) to the actual GDS main structure name via [GDS_MAIN_STR_NAME] (e.g. LSmitll_AND2T_v2p1).

• Layer name normalization for routing: DEFs may emit metal1, metal2, etc. ViPeR’s [Parameters] contains

  routingLayers = ["M1", "M3"]

  We now route:

  • metal1 → GDS layer 10 (M1)
  • metal2 → GDS layer 30 (M3)
    The PTL width comes from Parameters.PTLwidth (µm) and is converted to GDS units.

Why: Without these translations, no top‑level routing appeared (DEF layer names didn’t match the original hard‑coded mapping), and placeBias() couldn’t find x‑offsets for modern config files.

---

3) Via import decoupled from fill enable (critical bug fix)

Files: src/chipsmith/chipFill.cpp

• Previously, the via cell ViaM1M3 lived in [GDS_LOCATIONS] (with the fill cells) and was only imported when fill = true.
  But we always place via SREFs regardless of fill. Result: GDS had references to ViaM1M3 but no cell definition → vias appeared as “(ViaM1M3)” (empty) in KLayout.

• Fix: We always call importFill() (which imports everything in [GDS_LOCATIONS]) and guard only the placement step (placeFill()) with fillEnable.
  This guarantees the via cell definition is present even when fill is disabled.

Why: Ensures via geometry is visible in the final GDS without forcing users to enable metal fill.

---

4) Build fixes for MSYS2/MinGW‑w64

Files: include/gdscpp/gdsParser.hpp, src/gdscpp/gdsCpp.cpp (and any file using fixed‑width integer types)

• Added the missing header:

  #include <cstdint>

  to fix 'uint32_t'/'uint64_t' was not declared errors.
• (Optional) You can statically link libstdc++/libgcc if you need a standalone EXE (larger file). The dynamic build is smaller and loads the runtime DLLs from the toolchain.

Why: Current GCC requires <cstdint> when using fixed‑width integer types.

---

5) Quality‑of‑life

• Printed the units and grid summary during importData().
• Parsed the configFileUnits and gdsUnits parameters from chipsmith_config.toml to replace hardcoded geometry scaling parameters in chipFill.cpp.
• Improved consistency of console logging (clearly separates import, routing, biasing, write stages).
• Kept the original via placement logic (SREFs at DEF via sites) but ensured the via cell is always available (see §3).

---