Generate QLT IR for all bloq examples

[mpharrigan]

Build an on-disk library of verified Qualtran-L1 (.qlt) files from bloqs.
Given a (bloq, name) pair and a root directory, build_library_entry runs the
full L1 pipeline for that bloq:

1. Compiles it to a .qlt file (qualtran.l1.dump_l1).
2. Loads that .qlt file back into bloqs (qualtran.l1.load_module).
3. Executes the root bloq through the
   StandardQualtranArchitectureAgnosticVirtualMachine.

---

This is the basic infrastructure. Future PRs will investigate and/or fix the failure cases.

[gemini-code-assist]

Code Review

This pull request introduces tools and libraries to build and verify an on-disk library of Qualtran-L1 (.qlt) files from BloqExamples, including a new driver script, shell script, and core build logic. Feedback on the changes highlights three critical issues: first, the use of Unix-only signal APIs (setitimer and SIGALRM) in _library.py will cause crashes on Windows and in multi-threaded environments; second, the newly added __str__ methods on both Join and Split bloqs reference a non-existent num_bits attribute on QDType instead of num_qubits, which will raise an AttributeError when called.

[gemini-code-assist]

The signal.setitimer and signal.SIGALRM APIs are Unix-only and will raise an AttributeError on Windows. Additionally, signal.signal can only be called from the main thread; calling it from a background thread (e.g., in a multi-threaded test runner or worker pool) will raise a ValueError. To prevent hard crashes in these environments, we should check for the presence of these APIs and ensure we are running on the main thread before attempting to set a timeout.

    import threading
    if (
        not seconds
        or seconds <= 0
        or not hasattr(signal, 'SIGALRM')
        or not hasattr(signal, 'setitimer')
        or threading.current_thread() is not threading.main_thread()
    ):
        yield
        return