Fix substring check used instead of equality for header names

The `in` operator on `bytes` performs substring search, not equality.
`header[0] in b"cookie"` matches any header name that is a substring
of "cookie" (e.g. b"co", b"ok", b"e"), not just b"cookie" itself.

This means short header names that happen to be substrings of "cookie"
get incorrectly promoted to NeverIndexedHeaderTuple when their value
is under 20 bytes, potentially affecting HPACK compression behavior.

Changed both occurrences to use `==` for exact comparison:
- Line 91: cookie header check in _secure_headers
- Line 350: :method pseudo-header check in _reject_pseudo_header_fields

Author: bysiber

src/h2/utilities.py

@@ -88,7 +88,7 @@ def _secure_headers(headers: Iterable[Header],
     """
     for header in headers:
         assert isinstance(header[0], bytes)
-        if header[0] in _SECURE_HEADERS or (header[0] in b"cookie" and len(header[1]) < 20):
+        if header[0] in _SECURE_HEADERS or (header[0] == b"cookie" and len(header[1]) < 20):
             yield NeverIndexedHeaderTuple(header[0], header[1])
         else:
             yield header
@@ -347,7 +347,7 @@ def _reject_pseudo_header_fields(headers: Iterable[Header],
                 msg = f"Received custom pseudo-header field {header[0]!r}"
                 raise ProtocolError(msg)
 
-            if header[0] in b":method":
+            if header[0] == b":method":
                 method = header[1]
 
         else: