docs: combined 388 392

[NiKrause]

Addresses @alanshaw recommendations from PR #388

This branch combines PRs #388 and #392 and addresses all review comments from @alanshaw.

Changes

README updates:

• Removed mock fetch functions from @ucanto/client and @ucanto/transport READMEs
• Updated examples to show real server usage (via environment variables)
• Changed ed25519.Verifier.parse() to DID.parse() for service DIDs (as suggested)
• Added notes about using server-as-channel for testing, with links to server README

Test improvements:

• Converted packages/client/test/client.spec.js from mock fetch to server-as-channel pattern
• Tests now validate delegation chains (no bypass)
• Added @ucanto/server and @ucanto/validator to client devDependencies

Benefits

• Examples work with real servers (no mocks)
• Tests properly validate UCAN delegation chains
• Follows recommended patterns from the server package
• All tests passing with 100% coverage maintained

Related

• Addresses review comments from: docs: Fix outdated READMEs for @ucanto/client and @ucanto/transport #388
• Should complete work for: docs: verify and update README examples for latest releases #387 (which was respected in docs: Documentation and Testing Improvements fixing #387 #392)

[NiKrause]

Those combined PR's are now included in #409 which also includes the new ed25519 webcrypto PR #408 (including docs)

The idea is that the code snippets in the readme's have tests - so doc isn't claiming any old API.

[hannahhoward]

Honestly this is such a massive improvement I'm just gonna merge. Thanks so much.

[hannahhoward]

eek... @NiKrause totally missed that this appears to be failing typechecking -- can you investigate?

[NiKrause]

Yes, interestingly before merge when creating PR typechecking was ok. I will look into today.

[NiKrause]

• This PR docs: combined 388 392 #407 had only Validate PR title (pull_request_target) runs.
• Package workflows (pull_request) did not run on that head SHA.
• main was effectively unprotected / ruleset disabled, so missing checks didn’t block merge.

I created a new PR #410 where I
allowed running on push for all branches in all packages:

  push:
    branches:
      - '**'

So if a branch introduces a type/test regression in covered paths, those checks will now fail on that branch push, earlier than merge. It broadens detection.