Fix api_schemas/json_encoders by manually converting NaT values to empty Strings

makes import of datetime columns more robust by first checking
if value is null because strftime can't handle NaT values

use `isnull()` because it handles all NaN/None/NaT cases

freqtrade/rpc/api_server/api_schemas.py

@@ -1,4 +1,5 @@
 from datetime import date, datetime
+from pandas import isnull
 from typing import Any, Dict, List, Optional, Union
 
 from pydantic import BaseModel
@@ -355,7 +356,9 @@ class PairHistory(BaseModel):
 
     class Config:
         json_encoders = {
-            datetime: lambda v: v.strftime(DATETIME_PRINT_FORMAT),
+            datetime: lambda v: v.strftime(DATETIME_PRINT_FORMAT)
+            # needed for aslong NaT doesn't work with strftime
+            if not isnull(v) else "",
         }