NPUW: Serialize the rerank flag in LLMCompiledModel blobs

m_is_rerank was set at compile time from NPUW_TEXT_RERANK but not
included in serialize()/deserialize(), so a rerank model exported and
re-imported (e.g. via ov::cache_dir) lost the flag and skipped the
batched-scoring wrapper in create_sync_infer_request(). Serialize it
alongside m_is_embedding and bump NPUW_SERIALIZATION_VERSION since the
blob layout changed.

Author: dylanneve1

src/plugins/intel_npu/src/plugin/npuw/llm_compiled_model.cpp

@@ -1319,7 +1319,7 @@ void ov::npuw::LLMCompiledModel::serialize(std::ostream& raw_stream, const ov::n
             m_kvcache_desc.dim & m_kvcache_desc.max_generation_token_len & m_kvcache_desc.v_tensors_transposed_pre &
             m_kvcache_desc.v_tensors_transposed_gen & m_prefill_chunk_size & m_use_chunk_prefill & m_max_lora_rank &
             m_enable_prefix_caching & m_prefix_caching_block_size & m_prefix_caching_max_num_blocks & m_is_whisper &
-            m_eos_token_id & m_decomposed_sdpa_size & m_is_eagle & m_is_embedding & m_is_block_kv_cache;
+            m_eos_token_id & m_decomposed_sdpa_size & m_is_eagle & m_is_embedding & m_is_block_kv_cache & m_is_rerank;
 
         // Write config
         stream & m_cfg;
@@ -1539,7 +1539,7 @@ std::shared_ptr<ov::npuw::LLMCompiledModel> ov::npuw::LLMCompiledModel::deserial
             compiled->m_use_chunk_prefill & compiled->m_max_lora_rank & compiled->m_enable_prefix_caching &
             compiled->m_prefix_caching_block_size & compiled->m_prefix_caching_max_num_blocks & compiled->m_is_whisper &
             compiled->m_eos_token_id & compiled->m_decomposed_sdpa_size & compiled->m_is_eagle &
-            compiled->m_is_embedding & compiled->m_is_block_kv_cache;
+            compiled->m_is_embedding & compiled->m_is_block_kv_cache & compiled->m_is_rerank;
 
         // Deserialize config
         stream & compiled->m_cfg;

src/plugins/intel_npu/src/plugin/npuw/serialization.hpp

@@ -48,7 +48,7 @@ const constexpr ov::npuw::s11n::IndicatorType NPUW_LLM_COMPILED_MODEL_INDICATOR
 const constexpr ov::npuw::s11n::IndicatorType NPUW_GQA_COMPILED_MODEL_INDICATOR =
     {char{0x47}, char{0x51}, char{0x41}, char{0x43}, char{0x4d}, char{0x4f}};
 
-const constexpr char* NPUW_SERIALIZATION_VERSION = "0.27";
+const constexpr char* NPUW_SERIALIZATION_VERSION = "0.28";
 
 // Forward declaration
 namespace intel_npu {