Standardize soft-fail input interpolation across reusable lint workflows

[WilliamBerryiii]

Summary

The 8 reusable lint workflows introduced in PR #240 (#235) use three inconsistent patterns for consuming the soft-fail input parameter. This creates unnecessary maintenance burden and makes behavior harder to reason about.

Current State

Pattern	Workflows	Example
Direct injection in if:	shell-lint, terraform-lint, code-quality-lint, bicep-lint	if: failure() && inputs.soft-fail == true
Environment variable	docs-lint, yaml-lint	env: SOFT_FAIL: ${{ inputs.soft-fail }} then if: env.SOFT_FAIL != 'true'
Negated condition	powershell-lint, security-scan	if: inputs.soft-fail != true

Acceptance Criteria

• Pick one canonical pattern for soft-fail input consumption
• Apply the chosen pattern uniformly across all 8 reusable workflows
• All lint jobs in pr-validation.yml and main.yml continue to function correctly
• Verify with a test PR that both hard-fail and soft-fail modes work as expected

References

• PR build(ci): replace MegaLinter with per-tool GitHub Actions lint workflows #240 — MegaLinter replacement
• Issue ci(lint): replace MegaLinter with dedicated per-tool GitHub Actions workflows #235 — Parent issue