Value parser sometimes not reliable with exponents

[kermitt2]

In this example, the raw value looks good but the parsed value is not very exciting.

1001._0908.0054.pdf

[lfoppiano]

oh, maybe adding some simple post-processing would avoid such mistakes.

[lfoppiano]

I've implemented some rules in #103:

• if whatever contained into is numeric, it goes into <number>
• if <number> comes after <pow> or <exp> it's probably the exponent or part of it and it's concatenated to the previous
• if <base> contains e then the following <pow> should go into <exp>