align_matrix yields invalid matrix in combination with missing float "0"

If the option `align_matrix = true` is used, the following happens.

#### Before formatting

```julia
A = [1. 1; 1 -1]
```

#### After formatting

```julia
A = [1.01; 1 -1]
```

#### Expected result

```julia
A = [1.0 1; 1 -1]
```