Streaming union read from the lake and Fluss log table using a specified offset timestamp

### Search before asking

- [x] I searched in the [issues](https://github.com/apache/fluss/issues) and found nothing similar.

### Motivation

I found that we cannot consume the log table using hybrid splits (lake + log) when a specific timestamp is provided.
My intention is to perform a streaming read starting from a given timestamp, but part of the data may reside in the underlying lakehouse rather than the log layer.

### Solution

_No response_

### Anything else?

_No response_

### Willingness to contribute

- [x] I'm willing to submit a PR!

No source files or tests are named in the issue. Start by tracing how hybrid splits choose lake and log data when a timestamp is supplied, then identify the streaming read entry point and relevant tests. Done means a read from the specified timestamp can combine data from both the underlying lakehouse and the Fluss log table.

Written by the indexing model from the issue text.