[R] Throw helpful errors on bad object types in dplyr expressions

When users pass bad dots args to `mutate()` and other dplyr verbs (such that they could not succeed even if the data were pulled into R first), they should get immediate, informative errors.

An example:
```java

mtcars %>% Table$create() %>% mutate(c)
```
 

**Reporter**: [Ian Cook](https://issues.apache.org/jira/browse/ARROW-12093) / @ianmcook
#### PRs and other links:
- [GitHub Pull Request #9805](https://github.com/apache/arrow/pull/9805)

_{**Note**: *This issue was originally created as [ARROW-12093](https://issues.apache.org/jira/browse/ARROW-12093). Please see the [migration documentation](https://github.com/apache/arrow/issues/14542) for further details.*}

Start by reproducing the R example using Table$create() and mutate(c), then trace how dplyr expressions and dots arguments are handled for mutate() and other verbs. Done means bad object types that could not succeed after collecting into R fail immediately with helpful, informative errors.

Written by the indexing model from the issue text.