fuzz_cesr_import: panic inside cesride 0.6.4 indexer (deadly signal after ~3.75M execs)

[bordumb]

The scheduled fuzz job on main@36eb6d67 hit a genuine crash after ~3,751,844 executions of fuzz_cesr_import:

thread '<unnamed>' panicked at cesride-0.6.4/src/core/indexer/mod.rs:396:25
== ERROR: libFuzzer: deadly signal
artifact: crates/auths-keri/fuzz/artifacts/fuzz_cesr_import/crash-9e0a22d2513d2e3e80fb2809cd5d1e0b7c4555d9

• The panic is inside the cesride dependency (Indexer construction), not in auths code — but it is reachable from fuzz_cesr_import's entry (attachment/indexed-signature import), i.e. attacker-supplied CESR bytes can panic the process instead of returning a typed decode error. The import path should catch/ward this input class before handing it to cesride (parse, don't panic), or cesride needs the upstream fix.
• Unrelated to the witness write-path changes (PR Witness receipting write path: per-prefix KELs, client publication, KEL-anchored witness sets #403's two fuzz jobs passed on the identical Rust tree; this run's random corpus walk found the input). Reproduction: the crash artifact was written to the runner's ephemeral disk and the workflow uploads no artifacts, so re-run cargo +nightly fuzz run fuzz_cesr_import (see docs on the verifier strengthened battery for local setup) — at ~3.7M execs/coverage 1579 the input class should resurface quickly; then cargo fuzz tmin it.
• Follow-up worth taking while there: make the fuzz workflow upload fuzz/artifacts/** on failure so findings survive the runner.