build: make wallet npm scripts cross-platform

[Sikkra]

Summary

Makes the wallet npm workflow usable from Windows/PowerShell while preserving the existing build outputs:

• fixes npm test by removing the shell-specific quoted glob and allowing Windows checkout line endings in ESLint
• lets npm run local-server use package.json defaults without POSIX environment assignment syntax
• replaces the shell-only npm build path with a Node build driver that runs the same Babel, UglifyJS, html-minifier, and clean-css steps
• adds a dist-start wrapper that sets the configured dist folder without POSIX-only VAR=value syntax

This keeps build.sh available for existing Unix users but makes the package scripts work cross-platform.

Verification

• Before: npm test failed on Windows with No files matching the pattern "'src/*.js'"; after: npm test passes.
• Before: npm run build failed on Windows with '.' is not recognized; after: npm run build completes and writes bitgesell-wallet-dist.
• node --check scripts/build.js scripts/dist-server.js local-server.js
• local server smoke: PORT=18080 node local-server.js returned HTTP/1.1 200 OK for /.

Program reference: BitgesellOfficial/bitgesell#81

USDT-TRC20 payout address if accepted: TPwPFww7zxXFQ7zugo22gktQhckWVarRqi

[MyTH-zyxeon]

Review-assist note for maintainers: this PR addresses a real cross-platform gap in the wallet workflow. It removes POSIX-only package scripts and moves the dist build into Node so Windows/PowerShell users can run the same npm paths as Unix users.

The main validation I would want before merge is:

• compare the built index.html, app.min.js, and custom.min.css from the old Unix build.sh path against the new scripts/build.js output for functional parity
• run the package scripts on Windows/PowerShell and on a Unix shell to confirm the new wrappers did not regress existing contributors
• inspect the diff with whitespace ignored because .eslintrc and local-server.js appear to include CRLF churn; the semantic changes are much smaller than the raw diff
• confirm packageConfig.DIST_FOLDER resolution still writes outside the repo exactly as before, and cannot accidentally delete/copy an unintended directory

I do not see a fatal blocker, but I would treat build-output parity plus the CRLF/noise review as the key acceptance checks.