fix #432 - Object usage line numbers

[russHyde]

In #432 it was noted that the line-number for object-usage-linter
warnings were often incorrect. They were being reported relative to the
start of the defining function, rather than the start of the file.

- A test was added that throws an object-usage-lint on line-2 of the
function but line-5 of the file.

- Modified object-usage-linter.R so that it reported the line-number from the original file, rather than the line-number of the encasing function

This led to several of the knitr-format tests failing because lints were throwing in a different order. The original tests were misordered , and have been modified.

Briefly:
Each of the knitr-format tests had code blocks that look like this:

~~~~
a = 1
~~~~

~~~~
b <- function(x) {
  d = 1
}

~~~~

The lints should be ordered:
- assignment (`=` for a)
- local variable ‘d’ assigned but may not be used
- assignment (`=` for d)
- trailing blank line

I updated the tests and added the explicit line-numbers where the lints
should throw to test-knitr_formats.R

[miker985]

I can confirm this fixes the reported issue in #432

[russHyde]

I've added a NEWS entry and squashed down to a single commit. Would you be able to review this @jimhester so that I can look at #445 (which is a slightly different bug)