Support Continue on Error in CSV Reader

**Describe the bug**
Attempting to read malformed CSV records where the number of columns doesn't match the schema leads to an out-of-memory error.

**To Reproduce**
See minimal example Rust code below.

[arrow_bug.txt](https://github.com/apache/arrow-rs/files/12575141/arrow_bug.txt)

**Expected behavior**
The CSV reader should start re-start reading from the first newline/delimiter after the malformed row is encountered.

Start with the minimal Rust reproduction in arrow_bug.txt and trace the CSV reader while processing a row whose column count does not match the schema. Confirm the malformed row can be skipped and reading resumes at the first newline or delimiter without an out-of-memory error.

Written by the indexing model from the issue text.