feat: fault/chaos injection (LocalStack Chaos API parity)

[k-rooot]

Floci is a drop-in LocalStack Community replacement, but has no equivalent of LocalStack's Chaos API / AWS FIS. I'd like to add opt-in fault injection for resilience testing — forcing errors, throttling, or latency per service/operation so clients can exercise retry/timeout/recovery paths.

Why it fits Floci's design: it emits real AWS wire-protocol errors (reusing AwsException + AwsExceptionMapper), not invented responses — so "always implement the real AWS wire protocol" holds. It's default-off and changes nothing unless a fault is armed. The community already wants faithful AWS errors (cf. #1911, #1728) — this reuses the same serialization.

Proposed shape:

• a low-priority ContainerRequestFilter in core/common, after ServiceEnabledFilter/AwsProtocolClaimFilter, consulting a fault registry service/action/region → error|latency);
• a /_localstack/chaos admin endpoint to arm/clear faults (alongside the existing /_localstack/* surfaces).

Before I build: @hectorvent — would you accept this upstream, and any constraints on the control-endpoint shape or filter ordering? Happy to bring a tested PR + a POC branch.

[hectorvent]

@k-rooot Sorry for the delay, and thanks for the proposal.We already have an FIS service implemented, so the first step is reviewing where that implementation currently stands.

Would you be interested in digging into it and proposing ideas? Happy to guide you and help align them with Floci's future goals.
Thanks!