Files uploaded with an uppercase or mixed-case extension (e.g. `123.PDF`,
`report.Docx`) failed to be parsed because the readFile worker matches the
extension against a lowercase switch (`'pdf'`, `'docx'`, ...), so an
uppercase suffix fell through to the default branch and threw
"... is not supported".
Normalize the extension to lowercase once at the lowest common entry point
`readFileContentByBuffer`, instead of at each business-layer call site. This
keeps business-layer behavior unchanged and covers all callers
(chat upload, external url, dataset, local file) in a single place.
Adds regression tests covering upper/mixed-case extensions.
Fixes #6996
Co-authored-by: DadaVinqi <DadaVinqi@users.noreply.github.com>
