[Bug]: Resolution of Collection with unfilled markers edgecolors depends on whether colors are passed in constructor or later

Nobody has claimed this yet.

### Bug summary

In the example below, I construct two collections of unfilled markers via `scatter(..., marker="x")`. In one case I set the marker colors in the constructor, in the other case I set it immediately after via an explicit setter call.
I then set the collection's facecolor (it doesn't matter here for unfilled markers, but one can imagine this is generic code where the marker used actually comes from elsewhere and I just have `marker=user_specified` and here the user just happened to be `"x"`.
I then probe the collection's edgecolor (which is how the cross will actually be drawn). In the first case, the edgecolor actually switched to blue, i.e. it followed the facecolor. In the second case it stays (red, green), i.e. it followed the set_color call and ignored the later set_facecolor.

### Code for reproduction

```python
from pylab import *

coll = plt.scatter([0, 1], [0, 1], marker="x",
color=["r", "g"])
coll.set_facecolor("b")
print(coll.get_edgecolor())

coll = plt.scatter([0, 1], [0, 1], marker="x")
coll.set_color(["r", "g"])
coll.set_facecolor("b")
print(coll.get_edgecolor())
```

### Actual outcome

[[0. 0. 1. 1.]]
[[1. 0. 0. 1. ]
[0. 0.5 0. 1. ]]

### Expected outcome

Same output for both collections.

### Additional information

_No response_

### Operating system

macos

### Matplotlib Version

3.8

### Matplotlib Backend

agg

### Python version

3.11

### Jupyter version

ENOSUCHLIB

### Installation

git checkout

Start by running the provided reproduction with matplotlib 3.8 and inspect the scatter entry point plus the collection's set_color, set_facecolor, and get_edgecolor behavior. Compare both construction paths and add or update coverage so unfilled-marker collections produce the same edgecolor result after set_facecolor.

Written by the indexing model from the issue text.