Improve SBOM file filtering logic

[dorser]

Improve SBOM file filtering logic (prepare for SOURCE classification support)

Description

Today, when loading file hashes from an SPDX SBOM, Micromize filters files using:

• fileTypes = BINARY or APPLICATION
• Plus a path-based mitigation: loading files under common executable paths (/bin, /sbin, /lib, /lib64, etc.)

This is a defensive workaround because many executable scripts are currently classified by Syft as TEXT, making them indistinguishable from non-executable text files (config, docs, docs, etc.).

If Syft improves SPDX classification to emit SOURCE for shebang scripts (See: anchore/syft#4640), we should:

1. Update filtering logic to include:
   • BINARY
   • APPLICATION
   • SOURCE
2. Reduce reliance on path-based heuristics.
3. Avoid loading unnecessary non-executable text artifacts into enforcement maps.

Proposed direction

Short term (Done)

• Refactor filtering logic into a dedicated classification function.
• Make path-based rules explicit and isolated.

Medium term (after and if Syft change is available)

• Include SOURCE in the allowed file types.
• Optionally gate behavior behind a feature flag to maintain backward compatibility.

Long term

• Consider augmenting classification with executable-bit metadata if available from SBOM generators in the future.
• Re-evaluate whether path-based heuristics can be fully removed once classification quality improves.