fix(examples): use Session instance for region in train-inference-e2e (#6050)

The train-inference-e2e-example config cell read AWS_REGION off the
Session class (AWS_REGION = Session.boto_region_name). boto_region_name
is an instance @Property, so class access returns the property object
rather than a region string; the subsequent
boto3.Session(region_name=AWS_REGION) then fails with
'TypeError: expected string or bytes-like object, got property' and
PYTORCH_TRAINING_IMAGE is built with a bogus region.

Instantiate a Session and read the property off the instance.

Co-authored-by: hrehard <hrehard@amazon.com>

Author: Hannah14295

v3-examples/inference-examples/train-inference-e2e-example.ipynb

@@ -61,7 +61,8 @@
     "TRAINING_JOB_PREFIX = \"e2e-v3-pytorch\"\n",
     "\n",
     "# AWS Configuration\n",
-    "AWS_REGION = Session.boto_region_name\n",
+    "session = Session()\n",
+    "AWS_REGION = session.boto_region_name\n",
     "PYTORCH_TRAINING_IMAGE = f\"763104351884.dkr.ecr.{AWS_REGION}.amazonaws.com/pytorch-training:1.13.1-cpu-py39\"\n",
     "\n",
     "# Generate unique identifiers\n",