[hail] variation of hl.float/hl.int that returns null on invalid string

[iitalics]

fixes #7408

stacked on #7452

adds new toplevel hail functions:

• parse_float, parse_float32, parse_float64
• parse_int, parse_int32, parse_int64

all of the above accept only tstr expressions, and return missing for invalid numbers.

>>> hl.eval(hl.parse_int("-5.3"))
>>> hl.eval(hl.parse_int("asdf"))
>>> hl.eval(hl.parse_int("-5"))
-5
>>> hl.eval(hl.parse_float("-NaM"))
>>> hl.eval(hl.parse_float("-NaN"))
nan
>>> hl.eval(hl.parse_float("5.3"))
5.3

[tpoterba]

You're parsing twice here -- once for the check and once for the parse. That's not ideal, but I do see the reason why it's not easy to do it just once in generated code without boxing.

[iitalics]

yeah :I if we could catch exceptions in asm4s then that would be one nicer way to handle this.

[tpoterba]

oops, need to add to functions rst files

[tpoterba]

(search for delimit and add in those places)

[iitalics]

what the hell, bytecode verify errors from using locals? i saw another code-function creating locals so i assumed it was safe to do so, but apparently it isn't.

[danking]

I extremely like this. Eventually, I'd like to be able to write the table reading code in Python. That import_matrix_table only supports certain types is frustrating and it'd be easier to fix with more parsing functions like this!