Detect installer command failures at execution time

[camjac251]

Summary

• Replace all seven %ERRORLEVEL% comparisons with the runtime
  if errorlevel 1 form.
• Ensure failures from mkdir and mklink are observed immediately
  and routed through the installer's existing error path.
• Use the same error-checking idiom for the initial administrator check.

Root cause

cmd.exe expands %ERRORLEVEL% when it parses a parenthesized command
block. Six installer checks are inside if/else blocks, so they can
read the status from before the preceding mkdir or mklink command
runs. This can report OK and continue after an operation failed.

if errorlevel 1 evaluates the current command status when that
statement executes, so the check observes the operation immediately
before it.

Verification

• Reproduced the parsing behavior with Windows cmd.exe: after a
  command returned exit status 7, the %ERRORLEVEL% form printed
  old-missed, while if errorlevel 1 printed new-detected.
• git diff --check develop...HEAD
• Confirmed all seven checks use if errorlevel 1 and no
  %ERRORLEVEL% checks remain.
• Confirmed resources/install.bat retains CRLF line endings.
• Run the installer from an elevated Windows Command Prompt in a
  disposable installation directory.

Compatibility

• Installation paths, link targets, command ordering, and user-facing
  messages are unchanged.
• The existing success and :error control-flow paths are unchanged.
• The replacement uses standard cmd.exe syntax and does not require
  delayed environment-variable expansion.
• The top-level administrator check was not affected by block expansion,
  but now uses the same error-checking form as the remaining installer.