[FIX] upgrade num2words library to 0.5.13

As of version 0.5.8, the num2words library does not correctly convert
some numbers to words in Arabic. It erroneously appends the Arabic
equivalent of "one" to the translation of 1000 (one thousand), when it
must be implicit.

Expected output:

'ألف و مئتان و أربعة و ثلاثون  , ست و خمسون'

0.5.9:
>>> from num2words import num2words
>>> num2words(1234.56, lang="ar")
'واحد ألف  و مئتان و أربعة و ثلاثون  , ست و خمسون'

This bug has been fixed in version 0.5.13 of the library:

0.5.13:
>>> from num2words import num2words
>>> num2words(1234.56, lang="ar")
'ألف و مئتان و أربعة و ثلاثون  , ست و خمسون'

This commit updates the requirements.txt to use the version of the
library that produces the correct output.

opw-3756778

References:
+ savoirfairelinux/num2words#512

Author: oomsveta

requirements.txt

@@ -24,7 +24,7 @@ lxml==4.6.5 ; python_version <= '3.10'  # min version = 4.5.0 (Focal - with secu
 lxml==4.9.2 ; python_version > '3.10'
 MarkupSafe==1.1.1 ; python_version <= '3.10'
 MarkupSafe==2.1.2 ; python_version > '3.10'
-num2words==0.5.9
+num2words==0.5.13
 ofxparse==0.19; python_version <= '3.9'
 ofxparse==0.21; python_version > '3.9'  # (Jammy)
 passlib==1.7.4 # min version = 1.7.2 (Focal with security backports)