Test AI suggestion field grouping for file:// and percent-encoded document paths, and a history ending in a continued line

## What happens

Two parsing branches in `UttrflowPredictCapture` have never run:

- `Sources/UttrflowPredictCapture/FieldReading.swift:115-116`: `directory(of:)` for a document given as a `file://` URL, and for a path with percent-encoded characters (such as `%20`). Tests only use plain paths;
- `Sources/UttrflowPredictCapture/ShellHistory.swift:29`: `commands(in:)` when the file's last command continues with a trailing backslash and there is no newline after it.

Measured on origin/main `26d7bc1` with `swift test --enable-code-coverage` (5,384 tests, all passing) and `xcrun llvm-cov export` over the test binary. The regions named below have an execution count of 0: no test runs them, so a change that breaks them passes `make verify`.

## Why it matters

`directory(of:)` decides which documents share one suggestions corpus. If `file://` or `%20` handling breaks, every document in a folder whose name has a space gets its own empty corpus, and suggestions stop working there. The history case silently drops the last command someone ran when importing a shell history.

## How to reproduce

Run `swift test --enable-code-coverage`, then `xcrun llvm-cov show -instr-profile .build/debug/codecov/default.profdata .build/debug/UttrflowPackageTests.xctest/Contents/MacOS/UttrflowPackageTests Sources/UttrflowPredictCapture/FieldReading.swift` and look for the `0` counts on the lines named above.

## Acceptance criteria

- A `FieldReading` whose document is `file:///Users/example/My%20Project/notes.md` groups with one whose document is `/Users/example/My Project/other.md`.
- `ShellHistory.commands(in: "ls\ngit commit \\\n -m done\\")` keeps `ls` and the continued command.
- Each new test fails when the branch it covers is broken. Check this by hand before opening the PR: temporarily change the branch (flip the condition or return the other value), see the test fail, then restore it.

## Where to start

- `Tests/UttrflowPredictCaptureTests/FieldReadingTests.swift` (suite "Telling one field from another by what it publishes") and `Tests/UttrflowPredictCaptureTests/ShellHistoryTests.swift`.

Read `CONTRIBUTING.md` first. Iterate with `swift test --filter `, then run `make verify` before pushing. Only tests change; no product code needs to.

**Size:** S, under an hour.

Read CONTRIBUTING.md, then inspect the named suites in Tests/UttrflowPredictCaptureTests/FieldReadingTests.swift and ShellHistoryTests.swift. Run the filtered Swift tests first, adding coverage for file:// and percent-encoded paths plus a continued final shell-history command. Confirm each test fails when its branch is temporarily broken, then run make verify.

Written by the indexing model from the issue text.