[backport camel-4.18.x] CAMEL-24187: CamelContext.getEndpoint() never caches a uri containing a % character

[Croway]

Summary

Backport of #24805 to camel-4.18.x (excludes the camel-plc4x test).

• AbstractCamelContext.doGetEndpoint() normalizes the URI once for cache lookup, but on a cache miss passes the already-normalized URI to addEndpointToRegistry(), which normalizes it a second time. Since EndpointHelper.normalizeEndpointUri() is not idempotent for URIs containing % (each pass encodes %41 → %2541), the stored registry key never matches a later lookup key — causing getEndpoint() to create a brand-new Endpoint on every call instead of reusing the cached singleton.
• Adds a normalized-aware overload of getEndpointKey()/addEndpointToRegistry() so the doGetEndpoint() call site skips redundant re-normalization. Public APIs (hasEndpoint, addEndpoint, removeEndpoints) are unchanged.
• Affects any component whose endpoint URI contains a literal % and is re-resolved by string (pollEnrich/toD/recipientList).

Test plan

• New test DefaultEndpointRegistryTest#testGetEndpointIsCachedForUriContainingPercentCharacter verifies the fix using controlbus:route?routeId=RAW(%41)&action=status.

Claude Opus 4.6 on behalf of Croway

[gnodet]

LGTM — clean backport of #24805 to camel-4.18.x.

The AbstractCamelContext fix and DefaultEndpointRegistryTest are identical to the main
branch version. The plc4x reproducer test is correctly excluded (component may not be on 4.18.x).

Already reviewed the original PR in detail — the normalized flag correctly prevents
double-normalization of URIs containing %, which made NormalizedUri keys non-idempotent
(%41 → %2541 on second pass).

Claude Code on behalf of gnodet