fix(cli): prevent welcome() from re-triggering setup when global config exists

[Bernardxu123]

Problem

When running reasonix without arguments from a directory that lacks reasonix.toml, the setup wizard is re-triggered even though a valid global config (~/.config/reasonix/config.toml) exists and loads successfully.

Root Cause

In welcome() (internal/cli/cli.go), the setup trigger condition was:

if src == "" && isInteractive() {   // src = config.SourcePath()
    interactiveSetup("reasonix.toml")  // ← triggers setup wizard
}

SourcePath() checks for reasonix.toml in cwd first, then falls back to the global user config. When the user runs reasonix from a different directory (e.g., after cd), ``SourcePath()returns""` because there's no cwd-local file — even though `config.Load()` would succeed from the global path.

The chat entry path also unnecessarily required src != "":

if src != "" && cfgErr == nil && isInteractive() {  // ← blocks chat when src is empty
    return chatREPL(nil)
}

Fix

1. Move config.Load() before the setup-trigger check
2. Change trigger condition from src == "" to src == "" && cfgErr != nil — wizard only runs when no config loads from ANY source
3. Remove redundant src != "" guard on chat entry — cfgErr == nil is sufficient

// Before: always triggered setup when no cwd-local file
if src == "" && isInteractive() {

// After: only triggers setup when config.Load() also fails
if src == "" && cfgErr != nil && isInteractive() {

Testing

• go build ./... — compiles successfully
• go test ./internal/config/ — all tests pass
• go test ./internal/cli/ — 5 pre-existing i18n test failures unrelated to this change

Closes #2856

[esengine]

Good catch and a clean diagnosis — welcome() keying the setup trigger on SourcePath() == "" (no cwd-local file) instead of on whether config.Load() actually succeeds was exactly the bug. Your fix (load early, gate on cfgErr != nil, drop the redundant src != "" on the chat path) is the right shape.

The branch had drifted into a conflict with main-v2 (the interactiveSetup signature changed since you opened this), so I reapplied your change verbatim onto current main-v2 in #2906 with your authorship preserved. It'll land there.

Closing in favour of #2906 — thanks for the contribution!