Use blake3 instead of blake2 for a considerable performance increase

Based on some rough benchmarking performed on reasonably modern,
but not over the top laptop hardware (i7-8665U + PCIe3 NVMe SSD)
this results in raw disk io (+ tar overhead) becoming the performance
bottleneck instead of hashing rate.

main.py

@@ -1,11 +1,11 @@
 #!/usr/bin/env python3
 from sys import stdin
 from sys import exit as sysexit
-from hashlib import blake2b
 from io import BytesIO
 from subprocess import Popen, PIPE
 
 import click
+from blake3 import blake3
 
 from identicon import Identicon
 
@@ -86,10 +86,11 @@ def print_usage_and_exit():
 
 
 def get_digest(stream):
-    hasher = blake2b(digest_size=DIGEST_SIZE)
+    # pylint: disable=not-callable
+    hasher = blake3()
     while data := stream.read(BUF_SIZE):
         hasher.update(data)
-    return hasher.digest()
+    return hasher.digest(length=DIGEST_SIZE)