Fix the two failures blocking PhyslibAlpha and QuantumInfo

[Gabrielebattimelli]

Why

The forced run got PhyslibAlpha built and analysed, then failed the load for two separate reasons.

1. QuantumInfo had no .olean files (33 errors)

error while processing ('QuantumInfo', 'Regularized'): uncaught exception:
object file '.../QuantumInfo/Regularized.olean' does not exist

The lake cache restored from physlib-lake-<sha> — a build made back when only Physlib was compiled. Lake then considered the tree up to date, so QuantumInfo was never built. Build PhysLib reported "Build completed successfully (9396 jobs)" with zero Built QuantumInfo.* lines.

My previous attempt put the workflow hash in the cache key but left a bare physlib-lake- restore-key, which matched that older cache anyway — so the fix did nothing. Now the key is namespaced by the workflow hash first and the restore-key stays inside that namespace.

2. An anonymous declaration aborted the entire load

File ".../jixia/structs.py", line 69, in <genexpr>
    name[-1].startswith(p) for p in ["_", "eq_", "match_", "proof_"]
IndexError: list index out of range

jixia's is_internal indexes name[-1] without checking for an empty name. Wrapped once at the import so all four call sites are covered, rather than guarding only the one that happened to crash.

Verified

raw([])       -> IndexError   (reproduces the crash)
guarded([])   -> True
guarded(real) -> False
guarded(_int) -> True         (normal behaviour preserved)

Plus: module compiles, YAML parses.