Minimize allocations parsing Content-Type (#444)

For context, this function is run on every request within ServeHTTP, and
the mime parsing and now lowercasing, is a marginal, but non trivial
amount of memory allocations.

The expensive bit is hitting the mime.FormatMediaType path, when we
already have a canonical form.

This removes the most expensive part over arguable the most common cases
where there are no additional parameters on the Content-Type.

```
$ go test -bench '^BenchmarkCanonicalizeContentType$' -run '^$' .
goos: darwin
goarch: arm64
pkg: github.com/bufbuild/connect-go
BenchmarkCanonicalizeContentType/simple-10         	92344741	        12.85 ns/op	       0 B/op	       0 allocs/op
BenchmarkCanonicalizeContentType/with_charset-10   	 1744219	       693.8 ns/op	     424 B/op	       6 allocs/op
BenchmarkCanonicalizeContentType/with_other_param-10         	 1969113	       614.4 ns/op	     424 B/op	       6 allocs/op
PASS
ok  	github.com/bufbuild/connect-go	5.800s
```

---------

Co-authored-by: Akshay Shah <akshay@akshayshah.org>

protocol.go

@@ -315,12 +315,35 @@ func flushResponseWriter(w http.ResponseWriter) {
 	}
 }
 
-func canonicalizeContentType(ct string) string {
-	base, params, err := mime.ParseMediaType(ct)
-	if err != nil {
-		return ct
+func canonicalizeContentType(contentType string) string {
+	// Typically, clients send Content-Type in canonical form, without
+	// parameters. In those cases, we'd like to avoid parsing and
+	// canonicalization overhead.
+	//
+	// See https://www.rfc-editor.org/rfc/rfc2045.html#section-5.1 for a full
+	// grammar.
+	var slashes int
+	for _, r := range contentType {
+		switch {
+		case r >= 'a' && r <= 'z':
+		case r == '.' || r == '+' || r == '-':
+		case r == '/':
+			slashes++
+		default:
+			return canonicalizeContentTypeSlow(contentType)
+		}
+	}
+	if slashes == 1 {
+		return contentType
 	}
+	return canonicalizeContentTypeSlow(contentType)
+}
 
+func canonicalizeContentTypeSlow(contentType string) string {
+	base, params, err := mime.ParseMediaType(contentType)
+	if err != nil {
+		return contentType
+	}
 	// According to RFC 9110 Section 8.3.2, the charset parameter value should be treated as case-insensitive.
 	// mime.FormatMediaType canonicalizes parameter names, but not parameter values,
 	// because the case sensitivity of a parameter value depends on its semantics.
@@ -329,6 +352,5 @@ func canonicalizeContentType(ct string) string {
 	if charset, ok := params["charset"]; ok {
 		params["charset"] = strings.ToLower(charset)
 	}
-
 	return mime.FormatMediaType(base, params)
 }

protocol_test.go

@@ -27,8 +27,10 @@ func TestCanonicalizeContentType(t *testing.T) {
 		arg  string
 		want string
 	}{
+		{name: "uppercase should be normalized", arg: "APPLICATION/json", want: "application/json"},
 		{name: "charset param should be treated as lowercase", arg: "application/json; charset=UTF-8", want: "application/json; charset=utf-8"},
 		{name: "non charset param should not be changed", arg: "multipart/form-data; boundary=fooBar", want: "multipart/form-data; boundary=fooBar"},
+		{name: "no parameters should be normalized", arg: "APPLICATION/json;  ", want: "application/json"},
 	}
 	for _, tt := range tests {
 		tt := tt
@@ -38,3 +40,26 @@ func TestCanonicalizeContentType(t *testing.T) {
 		})
 	}
 }
+
+func BenchmarkCanonicalizeContentType(b *testing.B) {
+	b.Run("simple", func(b *testing.B) {
+		for i := 0; i < b.N; i++ {
+			_ = canonicalizeContentType("application/json")
+		}
+		b.ReportAllocs()
+	})
+
+	b.Run("with charset", func(b *testing.B) {
+		for i := 0; i < b.N; i++ {
+			_ = canonicalizeContentType("application/json; charset=utf-8")
+		}
+		b.ReportAllocs()
+	})
+
+	b.Run("with other param", func(b *testing.B) {
+		for i := 0; i < b.N; i++ {
+			_ = canonicalizeContentType("application/json; foo=utf-8")
+		}
+		b.ReportAllocs()
+	})
+}