This is stacked on #608, so the diff currently includes that PR until it lands.

node-cron already requires Node 20+, and v5 is the right time to consider whether keeping a separate CommonJS build is still worth it. Node 20.19.0 enabled require(esm) by default, so CommonJS consumers on the supported runtime can load the same ESM entry instead of needing a second build artifact.

This removes the CJS build and points the package at the ESM output only:

"exports": {
  ".": {
    "types": "./dist/node-cron.d.ts",
    "module-sync": "./dist/node-cron.js",
    "default": "./dist/node-cron.js"
  }
}

To keep that technically accurate, the Node engine floor moves from >=20 to >=20.19.0, which is the Node 20 release where require(esm) stopped requiring --experimental-require-module. Reference: https://nodejs.org/en/blog/release/v20.19.0

The package-size impact is meaningful. Measured with npm pack --dry-run --json on clean temp copies:

#608 base
package size:   62,816 bytes
unpacked size: 327,517 bytes
entries:       17

this PR
package size:   37,396 bytes
unpacked size: 169,332 bytes
entries:       10

That is 25,420 fewer packed bytes (40.5% smaller) and 158,185 fewer unpacked bytes (48.3% smaller).

The deleted output is the duplicated CJS surface: node-cron.cjs, daemon.cjs, _shared.cjs, their sourcemaps, and the CJS-only declaration file from #608. It also removes the Rollup replace plugin that only existed to rewrite ESM-specific paths for the CJS build.

Checked with:

npm run build
npm run typecheck
npm run lint
npm test
npm pack --dry-run --json
npx @arethetypeswrong/cli --pack . --profile esm-only --format table --no-color

I also smoke-tested the packed package with both import styles on a Node runtime where process.features.require_module === true:

import cron, { schedule } from "node-cron";

const cron = require("node-cron");

And typed CommonJS consumption in TypeScript nodenext mode:

import cron = require("node-cron");

cron.schedule("* * * * *", () => {});