Improve performance of number parsing

[radeusgd]

Currently the NumberParser relies on a really complicated Regex that is not as efficient as we want.

We want to modify it to use a custom solution that should be more efficient.

TODO:

• what are the goals exactly?
• do we have benchmarks that measure the performance, so that we can compare? Parsing a column of Text values to Integer? CSV reading?

[jdunkerley]

New NumberParser is in #11499
This ticket should add Java unit tests to it and expose the functionality in Standard.Base so we can parse with the same capabilities there as well.

https://github.com/jdunkerley/scratch-code/blob/main/scratch-java/src/test/java/uk/co/jdunkerley/scratch/parser/DoublesTest.java
https://github.com/jdunkerley/scratch-code/blob/main/scratch-java/src/test/java/uk/co/jdunkerley/scratch/parser/IntegersTest.java
https://github.com/jdunkerley/scratch-code/blob/main/scratch-java/src/test/java/uk/co/jdunkerley/scratch/parser/SeparatorsTest.java

[jdunkerley]

Better integration of problems with the new Parser:

• Propagate the message from NumberParseFailure in some way.
• Get more friendly error reporting it would be great to save somewhere (perhaps in a SeparatorParseResult) what character was encountered to cause the 'invalid separators'.

[jdunkerley]

Exponential Notation:
1E+3
1E ==> 1 with symbol E
1,000,000.456E6
Rule was (0<=x<10)
1,000,000.456E6 ==> 1,000,000.456 and stop at E
23E6 => 23000000
Allow in exponent mode a single decimal point separator and allow a wider range than before (i.e. 0 <= x < 1000).

Merged version will be just enough to get it working again not deeply tested.

[jdunkerley]

Add a benchmark for parsing a single column and parsing 300 columns.

[jdunkerley]

ToDos:

• Redundant separator spacing check.