Decimal Points Missed Entirely by New Version

**Describe the bug**

Decimal points are sometimes not read by the program despite being in the pdf text. I.e., it reads "1.5" as "15". The is a **new** bug, as version 0.7.2 was working correctly. The current version (0.10.1) as well as 0.7.3 both fail.

**Steps to reproduce the bug**

1. Download this [table](https://www.osce.org/files/f/documents/6/2/369741.pdf).
2. Install camelot: `pip install camelot-py==0.10.1`
3. Run

`camelot -p all -o "test-NEW.csv" -f csv -split -strip ".\n" lattice -scale 100 -copy v "369746.pdf"`

4. Install older version: `pip install camelot-py==0.7.2`
5. Run

`camelot -p all -o "test-OLD.csv" -f csv -split -strip ".\n" lattice -scale 100 -copy v "369746.pdf"`

7. Open test-NEW-page-1-table-1.csv and test-OLD-page-1-table-1.csv.

**Expected behavior**

Line 2 of test-OLD.csv is what we **should** have:

"SMM camera at Donetsk Filtration Station (15km N of Donetsk)","**0.5-1.5km**","S","Recorded","2","Projectile","From E to W","N/K","31-Jan","19:35"

Line 2 of test-NEW.csv is **misread**:
"SMM camera at Donetsk Filtration Station (15km N of Donetsk)","**05-15km**","S","Recorded","1","Projectile","From E to W","N/K","31-Jan","19:34"

(Note that the same thing happens to the column name located in the first row -- "No." is converted into "No".)

**PDF**

[PDF](https://www.osce.org/files/f/documents/6/2/369741.pdf)

**Environment**

- OS: Garuda Linux (Arch-based)
- Python version: 3.10.2
- Numpy version: 1.22.2-1
- OpenCV version: 4.5.5-3
- Ghostscript version: 9.55.0-4
- Camelot version: 0.10.1

Start by running the listed camelot CLI command against the linked PDF and compare test-NEW-page-1-table-1.csv with the older output from version 0.7.2. Trace the camelot command entry point and PDF table-extraction path; done means decimal points and the period in "No." are preserved in the generated CSV, matching the expected older output.

Written by the indexing model from the issue text.