Startup and discovering 1341 mods takes 0.8s. Too slow

[PyXiion]

Version

0.1.0

Operating System

Linux

Describe the bug

On fresh startup, mod_service.reload takes ~800ms to discover 1341 mods.
The steam_cmd provider accounts for ~617ms of that, with process_mods taking 401ms.

Relevant logs

path                                count         time     own time     cpu time    own cpu time
--------------------------------  -------  -----------  -----------  -----------  --------------
mod_service.reload                      1  0.788296     0.656566     1.25371         1.12259
├── discover                            1  0.656321     2.5539e-05   1.12235         2.5378e-05
│   └── discover                        1  0.656296     0.653179     1.12232         1.11922
│       ├── provider.core               1  0.00989277   0.00132502   0.0151471       0.00203503
│       │   ├── scan_dir                1  0.000638561  0.000638561  0.00146523      0.00146523
│       │   └── parse_xml               6  0.00792919   0.00792919   0.0116469       0.0116469
│       ├── provider.local              1  0.209673     0.00292594   0.320519        0.00543333
│       │   ├── scan_dir                1  0.110362     0.110362     0.204356        0.204356
│       │   └── process_mods            1  0.0963849    0.0963849    0.110729        0.110729
│       ├── provider.steam_cmd          1  0.616949     0.00404832   1.08361         0.00448865
│       │   ├── scan_dir                1  0.211628     0.211628     0.32224         0.32224
│       │   └── process_mods            1  0.401272     0.401272     0.756882        0.756882
│       ├── resolve_active              1  0.00308341   0.00308341   0.00306919      0.00306919
│       └── ctx.load                    1  3.3503e-05   3.3503e-05   3.3623e-05      3.3623e-05
├── diagnostics                         1  0.0260597    1.4537e-05   0.0254605       1.4137e-05
│   └── diagnostics.init                1  0.0260452    0.00036108   0.0254464       0.000366677
│       ├── ensure_databases            1  0.0209227    0.0209227    0.020968        0.020968
│       └── load_community_rules        1  0.00476139   0.00476139   0.00411169      0.00411169
├── startup_impact                      1  0.0359902    0.0359902    0.0615619       0.0615619
└── diagnostics.rebuild                 1  0.13173      0.00649301   0.13112         0.00646479
├── collect_active                      1  0.0010798    0.0010798    0.00107437      0.00107437
├── graph.build                         1  0.00420675   0.00420675   0.00420171      0.00420171
├── find_cycles                         1  0.000571936  0.000571936  0.000572694     0.000572694
└── check_mod                         750  0.119379     0.119379     0.118806        0.118806 

Additional context

No response

[PyXiion]

I could add SQLite (the best DB ever) cache based on mtime.

If I rewrite the service to C++, it could take only about 100-200ms (without cache).
But C++ should be deferred for now.

With SQLite I can target about 40-80ms with cache for discover