Improved null handling when converting from nu -> dataframe. (nushell…

…#13855)

# Description
Fixes: nushell#12726 and nushell#13185

Previously converting columns that contained null caused polars to force
a dtype of object even when using a schema.

Now:
1. When using a schema, the type the schema defines for the column will
always be used.
2. When a schema is not used, the previous type is used when a value is
null.

# User-Facing Changes
- The type defined by the schema we be respected when passing in a null
value `[a]; [null] | polars into-df -s {a: str}` will create a df with
an str dtype column with one null value versus a column of type object.
- *BREAKING CHANGE* If you define a schema, all columns must be in the
schema.

crates/nu_plugin_polars/src/dataframe/command/aggregation/count.rs

@@ -32,12 +32,9 @@ impl PluginCommand for ExprCount {
     }
 
     fn examples(&self) -> Vec<Example> {
-        // to add an example with a result that contains a null we will need to be able to
-        // allow null values to be entered into the dataframe from nushell
-        // and retain the correct dtype. Right now null values cause the dtype to be object
         vec![Example {
             description: "Count the number of non-null values in a column",
-            example: r#"[[a]; ["foo"] ["bar"]] | polars into-df 
+            example: r#"[[a]; ["foo"] ["bar"] [null]] | polars into-df 
                     | polars select (polars col a | polars count) 
                     | polars collect"#,
             result: Some(