fix cannot find issue when performing collect on an eager dataframe (#15577)

# Description
Performing a `polars collect` on an eager dataframe should be a no-op
operation. However, when used with a pipeline and not saving to a value
a cache error occurs. This addresses that cache error.

crates/nu_plugin_polars/src/dataframe/command/data/collect.rs

@@ -5,7 +5,9 @@ use crate::{
 };
 
 use nu_plugin::{EngineInterface, EvaluatedCall, PluginCommand};
-use nu_protocol::{Category, Example, LabeledError, PipelineData, Signature, Span, Type, Value};
+use nu_protocol::{
+    Category, Example, LabeledError, PipelineData, ShellError, Signature, Span, Type, Value,
+};
 
 #[derive(Clone)]
 pub struct LazyCollect;
@@ -78,13 +80,21 @@ impl PluginCommand for LazyCollect {
                 // This should just increment the cache value.
                 // We can return a value back without incrementing the
                 // cache value or the value will be dropped (issue #12828)
-                let _ = plugin.cache.get(&df.id, true)?;
+                let cv =
+                    plugin
+                        .cache
+                        .get(&df.id, true)?
+                        .ok_or_else(|| ShellError::GenericError {
+                            error: format!("Failed to get cached value {}", df.id),
+                            msg: "".into(),
+                            span: Some(call.head),
+                            help: None,
+                            inner: vec![],
+                        })?;
+                let df = NuDataFrame::from_cache_value(cv.value.clone())?;
 
                 // just return the dataframe, add to cache again to be safe
-                Ok(PipelineData::Value(
+                Ok(PipelineData::Value(df.into_value(call.head), None))
-                    df.cache(plugin, engine, call.head)?.into_value(call.head),
-                    None,
-                ))
             }
             _ => Err(cant_convert_err(
                 &value,