MCP Updates (#666)

* mcp: fix robot scenarios for empty-result, literal-select, list_registry

Closes #661 follow-up.

- Empty-result test now uses google.cloudkms.key_rings (canonical empty-table
  case in the mock backend), and extractQueryResults handles the nil
  SQLResult that PrepareResultSet emits for empty RowMap as a zero-row result
  instead of an extraction failure - that's why list_registry against a
  fresh registry was returning rc=2.
- Literal-select assertion escapes &{ so Robot Framework doesn't parse it
  as a dictionary-variable opener.
- list_registry test asserts only rc=0 and absence of extraction failure so
  it works against both the canonical (test/registry) and mocked-empty
  (test/registry-mocked) registry configurations CI exercises.
- extractQueryResults: initialise rv as empty slice, guard nil row at EOF,
  honour GetError(); pairs with the (rv, ok) return so empty results render
  as "**no results**" rather than failing extraction.

Co-authored-by: Jeffrey Aven <jeffreyaven@users.noreply.github.com>

* mcp updates

* lint fixes

* build fixes

* build fixes

* build fixes

---------

Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: Jeffrey Aven <jeffreyaven@users.noreply.github.com>

graphql enhancements (#658)

* graphql enhancements

* added graphql tests

Merge remote-tracking branch 'origin/main' into feature/mcp-pr2

restore-shell-readlines-semicolon (#645)

Summary:

- Restore semicolons to appropriate places in `readlines` buffer from `stackql shell` sessions.
- Fix multiline query handling.
- `--approot` passing in test fix, in support of `readlines` buffer checking. `readlines` check skipped for `docker` tests.
- Added robot test `Shell Session Multi Line Then Multi Statement`.
- Tested correct semicolon placement in enhanced robot test `Shell Session Multiple Statements Inline`.
- Skip `readline` verification in docker for now.

static-analysis-coherence (#644)

Summary:

- Synchronising with latest `any-sdk` lib artifact:
    - Supports static analysis.
    - Supports auto generated, contrived mock testing.
    - Powerful reporting tool.
    - Statistical feedback to human and machine developers.

test: add unit and integration tests for annotatedast (#612)

* test: add unit and integration tests for annotatedast

* rerun CI

* rerun lint CI

* Update package comment to include nolint directive

* Update package comment to allow testpackage linting

* Delete internal/stackql/astanalysis/annotatedast/c.out

* Delete internal/stackql/astanalysis/annotatedast/coverage.html

* Delete internal/stackql/astanalysis/annotatedast/coverage.out

* Delete cover.out

---------

Co-authored-by: Benevolent General Kroll Who cannot spell <82620104+general-kroll-4-life@users.noreply.github.com>

mac-os-runner-arm-by-default (#610)

broader-aws-support (#608)

Summary:

- Support for `aws.s3` parameter mobility.
- Support for fine grained host parameters.
- Support for `context` parameters, a convenient abstraction loosely aligned to `golang` `context` at this point.
- Added handler unit testing.
- Added context variable unit testing.

minimise-mcp (#596)

Summary:

- Add CI check for `go.sum` consistency.
- Updated high level design diagram.
- MCP server amendments:
    - Remove namespaced tools.
    - Tool descriptions tagged with future proofing and deprecations.
    - Removed robot tests for excised tools.

Merge pull request #594 from stackql/claude/fix-stackql-tests-01FaQ7Y…

…N6kjqRvqDwtymJdDR

Add window function (OVER clause) and CTE (WITH clause)