Timezone offset parsers accept out-of-range minute fields

Codex found this during an algorithm-defect audit of the `time` codebase.

Audited revision: `8706e568aafcd4562fde792374b94b735ae53a0a`.

## Summary

Both `%z` parsing and the ISO 8601 `timeOffsetFormat` accept numeric timezone offsets without validating their hour/minute components. A minute field of `60` is normalized into an extra hour instead of being rejected.

## Reproduction

```haskell
parseTimeM False defaultTimeLocale "%z" "+2360" :: Maybe TimeZone
-- Just +2400

formatParseM (timeOffsetFormat ExtendedFormat) "+23:60"
-- Just +2400

formatParseM (timeOffsetFormat ExtendedFormat) "+24:01"
-- Just +2401
```

## Expected

Malformed offsets should fail. In particular, the minute component must be in `00..59`; the ISO parser should also enforce the permitted overall ISO 8601 offset range.

## Actual

The parsers calculate `hours * 60 + minutes`, accepting invalid input as a different offset.

## Affected code

- `lib/Data/Time/Format/Parse/Instances.hs`: `readTzOffset`.
- `lib/Data/Time/Format/ISO8601.hs`: `timeOffsetFormat` (`digits2` plus unrestricted `isoMap`).

Start with `readTzOffset` in `lib/Data/Time/Format/Parse/Instances.hs` and `timeOffsetFormat` in `lib/Data/Time/Format/ISO8601.hs`, reproducing the `%z` and ISO examples from the issue. Trace how the hour and minute fields are converted, then verify that malformed offsets fail while valid offsets continue to parse within the permitted range.

Written by the indexing model from the issue text.