Add WMT translate datasets for 2017-19, with trivial extensibility to additional years.

PiperOrigin-RevId: 237342321

Author: adarob

tensorflow_datasets/testing/test_data/fake_examples/wmt19_translate/commoncrawl/commoncrawl.cs-en.cs

@@ -0,0 +1,3 @@
+běžím
+zmizel
+Plav

tensorflow_datasets/testing/test_data/fake_examples/wmt19_translate/commoncrawl/commoncrawl.cs-en.en

@@ -0,0 +1,3 @@
+I am running (cc)
+
+I am swimming (cc)

tensorflow_datasets/testing/test_data/fake_examples/wmt19_translate/commoncrawl/commoncrawl.de-en.de

@@ -0,0 +1,3 @@
+ich renne
+es verschwand
+ich schwimme

tensorflow_datasets/testing/test_data/fake_examples/wmt19_translate/commoncrawl/commoncrawl.de-en.en

@@ -0,0 +1,3 @@
+I am running
+
+I am swimming

tensorflow_datasets/testing/test_data/fake_examples/wmt19_translate/czeng/convert_czeng16_to_17.pl

@@ -0,0 +1,14 @@
+# This is a fake perl script to mimic the one written to filter CzEng 1.6 to
+# create CzEng 1.7. Our code just parses it to find the blocks that need to be
+# filtered out.
+
+use strict;
+
+my %bad = map { ($_, 1) } qw{
+2 3 5
+9 10 16
+};
+
+
+print STDERR "Done.\n";
+