fix: raise instead of segfaulting when OcTree/ColorOcTree.read() gets a bad path

[wkentaro]

OcTree.read() / ColorOcTree.read() segfaulted instead of raising when handed a missing or unreadable path (#68). octomap's C++ AbstractOcTree::read() returns NULL in that case; the wrappers assigned it straight to thisptr, so the next method call dereferenced NULL and took down the interpreter with SIGSEGV.

Now the wrappers NULL-check after read() and raise a Python exception: FileNotFoundError for a missing path, OSError for an unreadable/corrupt file, and ValueError for bad serialized bytes. Both OcTree and ColorOcTree go through one shared helper.

Test plan

• octomap.OcTree.read("/nonexistent.ot").getTreeType() now raises FileNotFoundError (exit 0) instead of exit 139
• regression tests added: test_read_missing_path_raises and test_read_unreadable_file_raises in tests/octomap_test.py and tests/color_octree_test.py
• uv run pytest tests/ — 40 passed
• uv run ruff check on the touched tests — clean

Closes #68

[wkentaro]

This was generated by AI during PR processing.

Verdict: recommend-merge

Rebased onto current main (post-#67) — main's changes overlapped tests/octomap_test.py, so I rebased to get CI against the real merged state; the rebase was conflict-free. CI is now fully green on the rebased commit and merge state is CLEAN.

Verified the fix end-to-end: rebuilt the extension and ran the new read-error tests (OcTree + ColorOcTree, missing path and corrupt file) — all pass. The NULL-pointer segfault is now turned into FileNotFoundError (missing path, via os.stat), OSError "failed to read" (corrupt-but-existing file), and ValueError (bad serialized bytes).

Code review found no correctness issues: _raise_read_error is correctly module-level, fsencode/fsdecode round-trips, the error branching matches the tests, and there's no double-free (__dealloc__ is guarded by if self.thisptr:).

Ready for your review and merge.