Move `std` toward qualified modules and shrink prelude

The current std design exposes many concrete operations in the global prelude. In a language with full-module type inference, bare identifiers are scarce and should ideally be reserved for generic, high-frequency operations. Because of that, we have long names like `string_starts_with`.

We should explore moving std toward a more qualified structure, for example:

- `string::trim`
- `string::starts_with`
- `string::ends_with`
- `string::slice`
- `array::slice`
- `array::append`

while keeping the prelude focused on generic/core vocabulary such as:

- `map`
- `filter`
- `iter`
- `next`
- `join`
- `split`

Goals:

- reduce pressure on the global unqualified namespace
- make concrete APIs easier to discover by domain
- keep short unprefixed names available for principled generic operations
- improve long-term std surface consistency

Notes:

- this likely requires richer std build/bootstrap machinery
- this is a language-surface design direction, not just an internal refactor

Open design questions:

- what belongs in the permanent prelude vs qualified std modules?
- how should native std modules and Ferlium prelude modules be composed during std construction?

No files or tests are named. Start by locating the existing standard-library construction, native std modules, and Ferlium prelude modules, then trace how they are composed during bootstrap. Done means the qualified-module and permanent-prelude boundary, along with the required build composition, has a documented and agreed design.

Written by the indexing model from the issue text.