Dictionary sometimes uses Ordinal hash code calculation instead of OrdinalIgnoreCase #44695

See comment at #44681 (comment). When too many collisions are seen in a Dictionary<string, ...>(StringComparer.OrdinalIgnoreCase) or HashSet<string>(StringComparer.OrdinalIgnoreCase), we'll accidentally fall back to StringComparer.Ordinal after re-bucketing entries.

This is a regression from netcoreapp3.1 and should be serviced for net 5.0.x.