feat!: FFI logging options in Go open config; remove StartLogs

[rkuris]

Adds WithLogPath and WithLogFilter functional options to initialize env_logger via FFI at database open time.

• Full RUST_LOG filter support via env_logger::Builder::parse_filters()
• Log path required; use /dev/stdout for stdout; no defaults
• Logging initialized before DB open; remains initialized even if open fails
• Global per-process logger; fails when already initialized
• Remove deprecated StartLogs and LogConfig from Go; keep fwd_start_logs called internally
• Add trace log on Db::new() and comprehensive TestLogging covering error/success cases

BREAKING CHANGE: StartLogs API removed; use WithLogPath/WithLogFilter options on New()

Why this should be merged

• We need this immediately to help with debugging the rust side
• Logging must be configured at open to enable future per-database log strategies.
• Aligns Go-side configuration with Rust env_logger semantics using RUST_LOG.
• Unblocks downstream integration (e.g., coreth) to set logging level and output programmatically.

Now we support any logging option as configured from the go side, which means we probably need some configuration options in coreth to set up logging the way we want.

How this works

• Go: New(path, WithLogPath("/dev/stdout"), WithLogFilter("info")) initializes logging before DB open.
• Rust FFI: LogArgs validates UTF-8; env_logger::Builder parses RUST_LOG filters and targets the provided path.
• Logging is global per-process; subsequent init attempts return an error.
• If DB open fails, logger remains initialized (explicitly documented).

How this was tested

• TestLogging (Go):
• Empty path: logging disabled, DB opens
• Invalid path: initialization fails with clear error
• Valid path + trace: log file created and non-empty; contains “Opening Firewood database …”
• Second init: fails with “attempted to set a logger after the logging system was already initialized”

Workspace checks

• cargo fmt
• cargo test --workspace --features ethhash,logger --all-targets
• cargo clippy --workspace --features ethhash,logger --all-targets -- -D warnings
• cargo doc --no-deps

[alarso16]

I don't think we can do this until logs are supported per-database. Otherwise, we can't even make multiple databases in tests (since coreth Firewood doesn't know it's a test at instantiation). If this is added for the next release, I think I would just have to remove logging all together

[demosdemon]

We shouldn't recommend use of /dev/stdout as it will cause interleaving of the stdout from the parent process -- even if the parent process is also implemented Rust sharing the same stdlib statics.

• Any synchronization in Go around printing to stdout will be bypassed.
• Because File::open("/dev/stdout") returns a new fd, any synchronization in Rust will also be bypassed.

[demosdemon]

We should wrap file with std::io::BufWriter to make using stdout slightly more copacetic (doesn't prevent interleaving but makes it harder if we only write whole lines). env_logger calls flush after writing out every event.

[rkuris]

I don't think we can do this until logs are supported per-database. Otherwise, we can't even make multiple databases in tests (since coreth Firewood doesn't know it's a test at instantiation). If this is added for the next release, I think I would just have to remove logging all together

We could still do this. If the database open fails because the logger is already initialized, just create the database with no logging. That would be a workaround until we implement per-database logging correctly.