To resolve an issue with quoted map keys not correctly parsing, this PR
changes how it progresses after scanning a quote token.

The old behaviour was:
1. scanner reaches a :
2. pull the characters out of the current scan context buffer and turn it into the map key token
3. set the previous indent column to the beginning of the buffer
So next time it gets to another key in the map, it'll check whether the indent was the same.

When scanning a quote:
1. Scan from the beginning to the end of the quote sequence
2. Add a token, send back to the lexer, and reset the scan context
In this scenario, it reaches the :, but nothing is in the scan context buffer because it just went ahead and added the quote token. The previous indent column doesn't get reset, so when it gets to the next map key it thinks the indent increased, which is invalid yaml.

The new behaviour is essentially to not reset the context after scanning a quote, and when reaching a : check the context tokens, and if it's a quote set the previous indent column to that

When scanning an escaped single quote `''` there was an off by one error
because while the local `idx` variable is increased, the context's `idx`
is not. This leads to the progress returned being off by one, so the
scanner gets stuck in the `ctx.next()` loop forever, unable to progress.

When we scan escape characters in quoted scalars, the local idx is
shifted but not the underlying context. This adds a column skip to
ensure the ctx progresses in sync with the local quote scanning.

`scanQuote` will add to the origin buffer for the purpose of creating
the quote token, but would never reset it. This caused a problem when
nothing would reset the buffer, i.e. when a quoted value was followed by
a newline and the next map key.

I had adjusted the `:` scanning logic in a way that didn't cover all
cases.