Registering custom udf in datafusion-cli could be easier

[ratmice]

Is your feature request related to a problem or challenge?

It would be nice if we could essentially call geodatafusion::register(...), and reuse
the basic datafusion-cli without a lot of copy/paste code or maintenance burden.

datafusion-contrib/geodatafusion#84

Describe the solution you'd like

I intend to post a PR where we export most of the current main_inner function, from the binary
into the datafusion-cli lib, that allows people customizing datafusion-cli with a default CliSessionContext to just call into the library for most of the configuration/argument parsing.

Describe alternatives you've considered

No response

Additional context

No response

[cschanhniem]

The approach of exporting main_inner from the binary into a library function is sound. A clean pattern used by other Rust CLIs (e.g., ripgrep, fd):

1. Export pub fn run_cli(args: Vec<String>, ctx_factory: impl Fn() -> SessionContext) from datafusion-cli/src/lib.rs
2. The factory closure lets callers register UDFs/UDAFs before the REPL starts
3. The current main.rs simply calls run_cli(std::env::args().collect(), DefaultSessionContextFactory)
4. External projects like geodatafusion call it with their own factory

// lib.rs
pub fn run_cli<F>(args: Vec<String>, ctx_factory: F) -> Result<()>
where
    F: Fn() -> SessionContext,
{
    let ctx = ctx_factory();
    // ... rest of CLI setup using ctx
}

This avoids duplicating the 300+ lines of CLI initialization (REPL, PRINT options, file dispatch) while giving full UDF/UDAF access. I'd be happy to review the PR.