Add LODR support to online and offline recognizers (#2026)

This PR integrates LODR (Level-Ordered Deterministic Rescoring) support from Icefall into both online and offline recognizers, enabling LODR for LM shallow fusion and LM rescore.

- Extended OnlineLMConfig and OfflineLMConfig to include lodr_fst, lodr_scale, and lodr_backoff_id.
- Implemented LodrFst and LodrStateCost classes and wired them into RNN LM scoring in both online and offline code paths.
- Updated Python bindings, CLI entry points, examples, and CI test scripts to accept and exercise the new LODR options.

sherpa-onnx/csrc/offline-rnn-lm.cc

@@ -83,11 +83,11 @@ class OfflineRnnLM::Impl {
 };
 
 OfflineRnnLM::OfflineRnnLM(const OfflineLMConfig &config)
-    : impl_(std::make_unique<Impl>(config)) {}
+    : impl_(std::make_unique<Impl>(config)), OfflineLM(config) {}
 
 template <typename Manager>
 OfflineRnnLM::OfflineRnnLM(Manager *mgr, const OfflineLMConfig &config)
-    : impl_(std::make_unique<Impl>(mgr, config)) {}
+    : impl_(std::make_unique<Impl>(mgr, config)), OfflineLM(config) {}
 
 OfflineRnnLM::~OfflineRnnLM() = default;