ts: (1) add more strict type check for id, name from option. (2) add more strict conversion for id, name from option.

(1) Fix some buggy display result in tooltip, like 2 value axis scatter, add add test cases for them.
(2) Buggy results in tooltip.renderModel: 'richText' for each cases. And add test cases for them.
(3) Add styles to renderMode: 'richText', make them the same as renderMode: 'html'.
(4) Make `tooltip.confine` default `true` if `tooltip.renderMode` is `richText`. And fix the positioning of it to avoid to overflow the container, which will be cut.
(5) Fix the `tooltip.formatter` callback in `renderMode: 'richText'`, previously the concatenating rich text by `params.marker.content` does not work.
(6) Refactor the implementation of `Series['formatTooltip']`:
Add a abstract layer named `TooltipMarkup` to describe the tooltip layout requirement for each series,
instead of lots of messy code to concatenate `html` and `richText` in each series.
`TooltipMarkup` is responsible for generating `html` or `richText` finally.
Add `TooltipMarkup` is also in charge of sort the tooltip by `valueAsc`, `valueDesc`, `seriesAsc`, `seriesDesc`.

[TEST_CASES]: test/new-tooltip.html
[PENDING]: Do we need to expose API (like `tooltipMarkup.ts#TooltipMarkupStyleCreator`) to `tooltip.formatter` callback, by which users can add richText style themselves. At present users is not able to do that.
But if we expose that kind of API, there is an issue:
should we use ZRender style (use `fill` to describe text color)
or ECharts label style (use `color` to describe text color).

(1) fix data-transform specify dimension by index throw error.
(2) some little refactor about source creation original brought by feature data transform.

fix: [data-transform] Make the comparison operation accurate in common edge cases. Support parser 'number'.

Each level will step with an interval. Make the final results more stable on each zoom level.

Add level to the third parameter