Implement zero-copy tokenization for Word, SingleQuotedString, and Whitespace

  Convert token string fields to use Cow<'a, str> to enable zero-copy tokenization
  for commonly used tokens:
  - Word.value: Regular identifiers and keywords now borrow from source
  - SingleQuotedString: String literals borrow when no escape processing needed
  - Whitespace: Single-line and multi-line comments borrow from source

Also add benchmark for measuring tokenization performance

Author: eyalleshem

.gitignore

@@ -18,4 +18,7 @@ Cargo.lock
 
 *.swp
 
-.DS_store
+.DS_store
+
+# dhat profiler output files
+dhat*.json

Cargo.toml

@@ -48,6 +48,7 @@ visitor = ["sqlparser_derive"]
 bigdecimal = { version = "0.4.1", features = ["serde"], optional = true }
 log = "0.4"
 recursive = { version = "0.1.1", optional = true}
+unicase = "2.7"
 
 serde = { version = "1.0", default-features = false, features = ["derive", "alloc"], optional = true }
 # serde_json is only used in examples/cli, but we have to put it outside
@@ -60,7 +61,12 @@ sqlparser_derive = { version = "0.4.0", path = "derive", optional = true }
 simple_logger = "5.0"
 matches = "0.1"
 pretty_assertions = "1"
+sysinfo = "0.30"
+dhat = "0.3.3"
+criterion = "0.5"  
 
 [package.metadata.docs.rs]
 # Document these features on docs.rs
-features = ["serde", "visitor"]
+features = ["serde", "visitor"] 
+
+

profile.json.gz

@@ -31,3 +31,7 @@ criterion = "0.7"
 [[bench]]
 name = "sqlparser_bench"
 harness = false
+
+[[bench]]
+name = "tokenize_bench"
+harness = false