Automatic Data Promotion from Int to Float for dt.replace function

- Describe what feature you would like to see implemented.
The ``replace`` function should automatically promote integers to floats, if a float is used in an integer column.

- If possible, give an example of how it may look in the code and what result
will be produced.

Example:

```python
df=dt.Frame({"Date":["8/4/2020","8/4/2020","8/4/2020","8/4/2020","8/4/2020"],
"Weight":[35,40,30,50,50],
"Subportfolio":["Portfolio1","Portfolio2","Portfolio3","Portfolio4","Portfolio5"],
"Portfolio Code":[1,1,1,2,2]})

df

Date Weight Subportfolio Portfolio Code
0 8/4/2020 35 Portfolio1 1
1 8/4/2020 40 Portfolio2 1
2 8/4/2020 30 Portfolio3 1
3 8/4/2020 50 Portfolio4 2
4 8/4/2020 50 Portfolio5 2

df.replace(1, 1.0)

# one scenario where a user might want to replace ints with floats is a column with nulls,
# where you have the mean of a list of numbers, but the mean is a float.
# Replacing the null entries in that column with the float will trigger an error as below :

---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
in
----> 1 df.replace(1, 1.0)

TypeError: Cannot replace integer value 1 with a value of type
```

References #2663