[REQ] Add wp.print_diagnostics() for build & runtime diagnostics #1221

Description

Add a wp.print_diagnostics() function (similar to jax.print_environment_info()) that prints a comprehensive snapshot of the Warp build and runtime environment — useful for debugging, bug reports, and CI logs.

The function prints human-readable output and returns a dict for programmatic access.

Information displayed

Software versions:

• Warp version (Python package, native library, warp-clang library)
• LLVM version (statically compiled into warp-clang — currently not exposed to Python)
• Python version, NumPy version, OS/platform
• Git commit hash (if set by CI)

CUDA & library info:

• CUDA Toolkit version, CUDA Driver version, NVRTC version
• CUDA compatibility mode, MathDx enabled/disabled
• libmathdx version (currently not exposed to Python)
• NanoVDB version (currently not exposed to Python)
• Host C++ compiler used to build native library (currently not exposed to Python)

Build flags:

• Debug mode (WP_ENABLE_DEBUG — function exists in native code but no Python binding)
• Verify FP (WP_VERIFY_FP — build-time define, no query function)
• Fast math (compiler flag only, no preprocessor define or query function)

Devices:

• Device list with alias, name, memory, architecture, mempool status

Example output

warp:              1.12.0.dev0
warp native:       1.12.0.dev0
warp-clang:        1.12.0.dev0 (LLVM 22.0.0)
numpy:             2.4.1
python:            3.12.3 (main, Jan 22 2026, 20:57:42) [GCC 13.3.0]
platform:          Linux-6.8.0-94-generic-x86_64

CUDA enabled:      True
CUDA Toolkit:      13.1
CUDA Driver:       13.1
NVRTC:             13.1
CUDA compat:       False
MathDx enabled:    True
libmathdx:         0.3.0
NanoVDB:           32.8.0
Host compiler:     GCC 13.3.0

Build flags:
  Debug:           False
  Verify FP:       False
  Fast math:       False

Devices:
  "cpu"       "CPU"
  "cuda:0"    "NVIDIA RTX PRO 6000 ..." (97 GiB, sm_100, mempool enabled)

Context

Currently there is no single function to dump all build and environment information. Users must piece together wp.__version__, wp.get_cuda_toolkit_version(), etc. Several pieces of build info — LLVM version, NanoVDB version, libmathdx version, host compiler version, and build flags (debug, verify_fp, fast_math) — are not exposed to Python at all, requiring new native C/C++ query functions and ctypes bindings.

This feature is modeled after jax.print_environment_info() and would be valuable for:

• Bug reports: Users can paste the output for complete environment context
• CI diagnostics: Quick verification of build configuration
• Debugging: Identifying mismatched versions or unexpected build flags

Implementation scope

• Native C/C++ changes: ~6 new WP_API functions across warp.h, warp.cpp, warp.cu, volume.cpp, clang.cpp
• Build system: Add WP_FAST_MATH preprocessor define to build_dll.py
• Python bindings: ctypes registrations, Runtime methods, module-level functions in context.py
• Public API: wp.print_diagnostics() + individual getters exported via __init__.py
• Tests: New test file for the function and individual getters