Detect numpy types when creating tables

[simonw]

Inspired by #8

[simonw]

Rough sketch:

+try:
+    import numpy
+except ImportError:
+    numpy = None
+
 Column = namedtuple(
     "Column", ("cid", "name", "type", "notnull", "default_value", "is_pk")
 )

@@ -70,6 +79,22 @@ class Database:
             datetime.time: "TEXT",
             None.__class__: "TEXT",
         }
+        # If numpy is available, add more types
+        if numpy:
+            col_type_mapping.update({
+                numpy.int8: "INTEGER",
+                numpy.int16: "INTEGER",
+                numpy.int32: "INTEGER",
+                numpy.int64: "INTEGER",
+                numpy.uint8: "INTEGER",
+                numpy.uint16: "INTEGER",
+                numpy.uint32: "INTEGER",
+                numpy.uint64: "INTEGER",
+                numpy.float16: "FLOAT",
+                numpy.float32: "FLOAT",
+                numpy.float64: "FLOAT",
+                numpy.float128: "FLOAT",
+            })

[simonw]

Need to test this both with and without numpy installed.