[Vulcan] Clarify workbook text encoding

[iliaal]

Summary

• Put text-encoding guidance in its own section, keeping the sheet-deletion explanation together.
• Make the UTF-8 default explicit and clarify that a selected Latin-1 locale also controls returned cell text, including after a workbook reload.
• Qualify the existing surrogate-fallback sentence with its required UTF-8 locale availability and link to the existing API reference.

Evidence

• The API reference documents input/output encoding and preserved locale across reloads.

• Existing regression 248 covers UTF-8 input/output by default.

• Existing regression 253 checks Latin-1 cell output after load() and loadFile(), for XLS and XLSX.

• The fallback requires an installed suitable UTF-8 locale.

Validation

• Verified all 300 current-default repository blobs and file modes before editing; only README.markdown changes.
• Strict native compilation with -Wall -Wextra -Wimplicit-fallthrough -Werror: pass, zero diagnostics.
• Full existing PHPT suite on the final documentation snapshot: 254 passed, 7 skipped, 0 failed, 0 warned.
• All 12 PHP API-reference files lint clean; whitespace check passes.
• Actually restored the baseline README: documentation-contract assertions failed as expected; restored the proposed README and all seven assertions passed. This checks documentation content, not runtime behavior.
• PHP 8.4.26 NTS, LibXL 5.2.0.1 trial, Linux. The compile reused an existing configure header; no fresh phpize/configure was run. Licensed/benchmark, ZTS, and Windows-only tests account for the skips. Other PHP/LibXL versions and Windows were not exercised locally.

Scope

Documentation only; existing regression coverage already establishes the behavior, so there is no new runtime test or implementation change. Released changelog entries remain frozen: the 2.9.1 getter-encoding wording is intentionally left unchanged.