perf: implement SIMD-accelerated event processing and optimize streaming performance

- Add SIMD utilities for fast byte array comparison and discriminator matching
- Optimize event processor with batch processing and memory pool
- Refactor global state management with concurrent data structures
- Remove deprecated batch processing module
- Enhance metrics collection with reduced overhead
- Improve parser efficiency across all protocol implementations
- Add performance benchmarking dependencies (criterion, wide)
- Update documentation and examples for new architecture

Performance improvements:
- SIMD-accelerated byte operations for instruction parsing
- Concurrent HashMap (DashMap) for better multi-threading
- Optimized memory allocation patterns
- Reduced lock contention in event processing pipeline

Breaking changes: Removed batch.rs module, updated parser interfaces

src/streaming/event_parser/protocols/block/block_meta_event.rs

@@ -1,3 +1,5 @@
+use std::borrow::Cow;
+
 use crate::impl_unified_event;
 use crate::streaming::event_parser::common::{types::EventType, EventMetadata};
 use borsh::BorshDeserialize;
@@ -20,8 +22,7 @@ impl BlockMetaEvent {
         program_received_time_us: i64,
     ) -> Self {
         let metadata = EventMetadata::new(
-            format!("block_{}_{}", slot, block_hash),
-            "".to_string(),
+            Cow::Borrowed(""),
             slot,
             block_time_ms / 1000,
             block_time_ms,