`ByteStream::from_file` is _much_ slower than pre-reading the data into memory first

[rcoh]

[rmliddle]

I believe this is at least in part because we are performing unbuffered reads: https://github.com/smithy-lang/smithy-rs/blob/main/rust-runtime/aws-smithy-types/src/byte_stream/bytestream_util.rs#L264

Wrapping this file in a BufReader before performing the take appears to somewhat improve throughput