Make some interpreter errors a bit nicer

[fingolfin]

Previously, we produced errors like this:

gap> function() 123; end;
Syntax error: end expected
function() 123; end;
           ^^^

This now is slightly more helpful and reads as follows.

gap> function() 123; end;
Syntax error: while parsing a function: statement or 'end' expected
function() 123; end;
           ^^^

Similar adjustments are made for loops, atomic blocks and if statements.

Resolves #1117

[coveralls]

Coverage remained the same at 85.145% when pulling a086dc1 on fingolfin:mh/nicer-errors into b4836bb on gap-system:master.

[fingolfin]

reading this again today, perhaps a semicolon would make this a bit clearer, as here (note that the "expected" is added automatically at the end, so we have to contort ourselves a bit to fit the sentence structure in there)

Suggested change

	Syntax error: inside a function, statement or 'end' expected in stream:5
	Syntax error: inside a function; statement or 'end' expected in stream:5

Thoughts?

[ChrisJefferson]

While ; might be more mathematically correct, I feel that in code I always read it as "end of statement".

How about "Syntax error while parsing a function: statement or 'end' expected" (so push the new stuff before the semicolon)

[fingolfin]

The "Syntax error:" prefix and "expected" suffix are fixed, I can only fill in stuff between them, so the best I can offer w/o deeper changes is this:

Syntax error: while parsing a function: statement or 'end' expected

[wilfwilson]

What do you think about either of these options, which break the error into two sentences with a full stop? (It also adds commas around the "or" option; the difference between the two options is the presence of the word "was")

Syntax error: inside a function. A statement, or 'end', was expected in stream:5

Syntax error: inside a function. A statement, or 'end', expected in stream:5

However I don't feel particularly strongly about any of the suggestions put forth in this PR/discussion, so I don't really mind.

[fingolfin]

Now updated

[fingolfin]

@wilfwilson sorry, I did not actually see your comment before I updated this PR, i.e., I did not deliberately ignore it! Your variant is also fine by me. I'll think about it again tomorrow, after some sleep. Perhaps others have an opinion, too

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@6424be9). Click here to learn what that means.
The diff coverage is 100%.

@@            Coverage Diff            @@
##             master    #3372   +/-   ##
=========================================
  Coverage          ?   85.17%           
=========================================
  Files             ?      697           
  Lines             ?   344559           
  Branches          ?        0           
=========================================
  Hits              ?   293487           
  Misses            ?    51072           
  Partials          ?        0

Impacted Files	Coverage Δ
src/read.c	94.4% <100%> (ø)

[fingolfin]

I think I'll just merge this as is: it is an improvement over the current status quo, and a truly better message will require more flexibility in the code which generates those messages (to not force the trailing "expected" etc.).