Epic: js-stellar-sdk modernization

[Ryang-21]

Goal

Modernize the js-stellar-sdk toolchain and runtime dependencies to align with current JavaScript ecosystem standards. The end state is an ESM-first library with zero HTTP dependencies, a Rollup-based build pipeline, and a consolidated codebase that absorbs the migrated TypeScript js-stellar-base.

Motivation

• axios and feaxios add bundle weight and maintenance surface that modern runtimes no longer require — fetch is available in Node 18+, all evergreen browsers, Deno, Bun, and Cloudflare Workers.
• CommonJS constructs block ESM consumers, tree-shaking, and modern bundler optimizations.
• The existing Webpack/Babel pipeline is slow, complex, and produces suboptimal ESM output. Rollup is the standard for library builds.
• Keeping js-stellar-base as a separate package forces dual releases, cross-repo coordination, and redundant CI. Consolidating into a single repo simplifies versioning, typing, and contributor onboarding.

Success Criteria

• No runtime HTTP dependencies in package.json.
• "type": "module" with working dual ESM/CJS exports map.
• Rollup produces ESM, CJS, and browser bundles; published package installs and resolves cleanly across Node, browser, and bundler consumers.
• js-stellar-base lives in this repo; downstream @stellar/stellar-sdk consumers depend on a single package.
• All existing tests pass; published artifacts import successfully in a fresh Node ESM, Node CJS, and bundled browser project.

Tasks (in order)

• Modernization 1/4: Replace axios and feaxios with a native fetch client #1384 — Replace axios and feaxios with a native fetch client
• Modernization 2/4: Remove CJS constructs and make the library ESM compliant #1385 — Remove CJS constructs and make the library ESM compliant
• Modernization 3/4: Move build tooling to Rollup #1386 — Move build tooling to Rollup
• Modernization 4/4: Move migrated TypeScript js-stellar-base into the repo #1387 — Move migrated TypeScript js-stellar-base into the repo

Out of Scope

• API redesign of public SDK surfaces (Horizon, RPC, contract, webauth).
• Migrating off Vitest.
• Changes to the CLI (src/cli/) beyond what the ESM/Rollup migration requires.
• Deprecating or removing existing public exports.

Sequencing Notes

Each task should ship as its own release (minor or major as semver dictates) so regressions can be bisected cleanly. Tasks 1 and 2 can overlap in review but should land in that order to keep diffs tractable. Task 4 must land last because absorbing base multiplies the surface affected by any earlier refactor.