Node not parsed correctly by xgb.model.dt.tree if split criteria is "<Inf"

Hi,

I noticed that in the table produced by `xgb.model.dt.tree` some nodes have NA in the "Feature" column, e. g.:
```
Tree Node ID Feature Split Yes No Missing Quality Cover
0 55 0-55 NA 0-NA 0-NA 0-NA NA NA
```
Afer looking through the code of the function, I think the issue is due to a regex that parses the text dump. Here's the raw text:

`55:[f279