Runtime: use SmallVector instead of std::vector to avoid memory allocations in most cases.

This dramatically reduces the number of needed malloc calls.
Unfortunately I had to add the implementation of SmallVectorBase::grow_pod to the runtime, as we don't link LLVM. This is a bad hack, but better than re-inventing a new SmallVector implementation.

SR-10028
rdar://problem/48575729

stdlib/public/runtime/ProtocolConformance.cpp

@@ -162,7 +162,7 @@ template<>
 const WitnessTable *
 ProtocolConformanceDescriptor::getWitnessTable(const Metadata *type) const {
   // If needed, check the conditional requirements.
-  std::vector<const void *> conditionalArgs;
+  SmallVector<const void *, 8> conditionalArgs;
   if (hasConditionalRequirements()) {
     SubstGenericParametersFromMetadata substitutions(type);
     bool failed =
@@ -626,7 +626,7 @@ swift::_searchConformancesByMangledTypeName(Demangle::NodePointer node) {
 
 bool swift::_checkGenericRequirements(
                       llvm::ArrayRef<GenericRequirementDescriptor> requirements,
-                      std::vector<const void *> &extraArguments,
+                      SmallVectorImpl<const void *> &extraArguments,
                       SubstGenericParameterFn substGenericParam,
                       SubstDependentWitnessTableFn substWitnessTable) {
   for (const auto &req : requirements) {