[Enhance]: Relax collection, field, and document ID name validation

### Affected Component

DB

### Current Behavior

The current name validation rules are very restrictive:

```c++
const std::regex COLLECTION_NAME_REGEX("^[a-zA-Z0-9_-]{3,64}$");

const std::regex FIELD_NAME_REGEX("^[a-zA-Z0-9_-]{1,32}$");

const std::regex DOC_PK_REGEX("^[a-zA-Z0-9_!@#$%+=.-]{1,64}$");
```

I don't see enough reasons to have such restrictions.

### Desired Improvement

Lifting or relaxing these restrictions, or making them configurable, for:

- collection names
- field names
- document primary keys / IDs

A more permissive validation rule, or a user-configurable validator, would make Zvec easier to adopt in heterogeneous environments.

A more clear and user-friendly message indicating what is not allowed.

### Impact

- Better compatibility with existing datasets and external identifiers
- Easier migration from other storage systems
- Reduced need for application-side name rewriting or ID normalization
- Less friction when integrating Zvec with multi-language tooling