[Bug][Jupyter Widget] Keplergl can't recognized converted float64 column

**Describe the bug**
Keplergl can't recognized converted `float64` column

**To Reproduce**
Steps to reproduce the behavior:
download the dataset from kaggle [=> download link](https://www.kaggle.com/mohitkr05/global-significant-earthquake-database-from-2150bc)
```
df = pd.read_csv('Worldwide-Earthquake-database.csv')
df['LONGITUDE'] = pd.to_numeric(df.LONGITUDE, errors='coerce')
df['LATITUDE'] = pd.to_numeric(df.LATITUDE, errors='coerce')
df['FOCAL_DEPTH'] = pd.to_numeric(df.FOCAL_DEPTH, errors='coerce')
df['EQ_MAG_MW'] = pd.to_numeric(df.EQ_MAG_MW, errors='coerce')

kepler_map = keplergl.KeplerGl()
kepler_map.add_data(data=df, name="earthquakes")
kepler_map
```
As you can see, `FOCAL_DEPTH` and `EQ_MAG_MW` are formated as `string` instead `float64`

**Expected behavior**
Keplergl detect automatically column types based on `pandas` dataframe dtypes.

**Environment (please complete the following information):**
- Python version: 3.7.10
- keplergl version : 0.3.2
- jupyterlab version : 3.1.9