Parse whole words only

Question/Bug

When using search_dates, it picks partial letters from neighboring words as well thus forming an invalid token and wrong datetime objects. For eg:

sample_string = 'Bubble -58.5 06 Mar 2009 in need of -43.4 30 Oct 1974 also contributed for -17.7 26 Dec 2018 '
en_dates = search_dates(sample_string, languages=['en'],settings={'STRICT_PARSING': True})

Output

[('5 06 Mar 2009 in', datetime.datetime(2009, 3, 5, 0, 0)),
 ('4 30 Oct 1974', datetime.datetime(1974, 10, 4, 0, 0)),
 ('7 26 Dec 2018', datetime.datetime(2018, 12, 7, 0, 0))]

It also picked

1. letter 5 from -58.5 to 06 Mar 2009 forming '5 06 Mar 2009 in'
2. letter 4 from -43.4 to 30 Oct 1974 forming '4 30 Oct 1974'
3. letter 7 from -17.7 to 26 Dec 2018 forming '7 26 Dec 2018'

Either include the whole word or exclude it. Just including partial numbers/letters from previous words makes it an invalid token and wrong DateTime objects.

Expected Output

[('06 Mar 2009 in', datetime.datetime(2009, 3, 5, 0, 0)),
 ('30 Oct 1974', datetime.datetime(1974, 10, 4, 0, 0)),
 ('26 Dec 2018', datetime.datetime(2018, 12, 7, 0, 0))]

Start by reproducing the reported example through the search_dates entry point and inspect how neighboring numeric characters are incorporated into candidate date tokens. Done means partial characters from adjacent words are excluded while the three expected date matches and datetime values are preserved; add a regression test for the example.